spark-nlp 6.0.2__py2.py3-none-any.whl → 6.0.4__py2.py3-none-any.whl

{spark_nlp-6.0.2.dist-info → spark_nlp-6.0.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: spark-nlp
-Version: 6.0.2
+Version: 6.0.4
 Summary: John Snow Labs Spark NLP is a natural language processing library built on top of Apache Spark ML. It provides simple, performant & accurate NLP annotations for machine learning pipelines, that scale easily in a distributed environment.
 Home-page: https://github.com/JohnSnowLabs/spark-nlp
 Author: John Snow Labs
@@ -102,7 +102,7 @@ $ java -version
 $ conda create -n sparknlp python=3.7 -y
 $ conda activate sparknlp
 # spark-nlp by default is based on pyspark 3.x
-$ pip install spark-nlp==6.0.2 pyspark==3.3.1
+$ pip install spark-nlp==6.0.4 pyspark==3.3.1
 ```
 
 In Python console or Jupyter `Python3` kernel:
@@ -168,7 +168,7 @@ For a quick example of using pipelines and models take a look at our official [d
 
 ### Apache Spark Support
 
-Spark NLP *6.0.2* has been built on top of Apache Spark 3.4 while fully supports Apache Spark 3.0.x, 3.1.x, 3.2.x, 3.3.x, 3.4.x, and 3.5.x
+Spark NLP *6.0.4* has been built on top of Apache Spark 3.4 while fully supports Apache Spark 3.0.x, 3.1.x, 3.2.x, 3.3.x, 3.4.x, and 3.5.x
 
 | Spark NLP | Apache Spark 3.5.x | Apache Spark 3.4.x | Apache Spark 3.3.x | Apache Spark 3.2.x | Apache Spark 3.1.x | Apache Spark 3.0.x | Apache Spark 2.4.x | Apache Spark 2.3.x |
 |-----------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|--------------------|
@@ -198,7 +198,7 @@ Find out more about 4.x `SparkNLP` versions in our official [documentation](http
 
 ### Databricks Support
 
-Spark NLP 6.0.2 has been tested and is compatible with the following runtimes:
+Spark NLP 6.0.4 has been tested and is compatible with the following runtimes:
 
 | **CPU**            | **GPU**            |
 |--------------------|--------------------|
@@ -215,7 +215,7 @@ We are compatible with older runtimes. For a full list check databricks support
 
 ### EMR Support
 
-Spark NLP 6.0.2 has been tested and is compatible with the following EMR releases:
+Spark NLP 6.0.4 has been tested and is compatible with the following EMR releases:
 
 | **EMR Release**    |
 |--------------------|

{spark_nlp-6.0.2.dist-info → spark_nlp-6.0.4.dist-info}/RECORD

@@ -3,16 +3,17 @@ com/johnsnowlabs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,
 com/johnsnowlabs/ml/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 com/johnsnowlabs/ml/ai/__init__.py,sha256=YQiK2M7U4d8y5irPy_HB8ae0mSpqS9583MH44pnKJXc,295
 com/johnsnowlabs/nlp/__init__.py,sha256=DPIVXtONO5xXyOk-HB0-sNiHAcco17NN13zPS_6Uw8c,294
-sparknlp/__init__.py,sha256=infpZcS--eZOsSSMengyAGWuh2IBDcKB9fvjQqQ6Keo,13814
+sparknlp/__init__.py,sha256=4IcJQhg7fuPsgeY0yoV96ZJPK_LVPdXnH3cl_azT7eU,13814
 sparknlp/annotation.py,sha256=I5zOxG5vV2RfPZfqN9enT1i4mo6oBcn3Lrzs37QiOiA,5635
 sparknlp/annotation_audio.py,sha256=iRV_InSVhgvAwSRe9NTbUH9v6OGvTM-FPCpSAKVu0mE,1917
 sparknlp/annotation_image.py,sha256=xhCe8Ko-77XqWVuuYHFrjKqF6zPd8Z-RY_rmZXNwCXU,2547
 sparknlp/functions.py,sha256=4cVRyBjlF1YttcMNs5z7gf9NPW7q9qzGb5KOf44Phgs,12120
 sparknlp/upload_to_hub.py,sha256=toULNLeXK3MmTpmza9dR8R8od2QJEw1eTgBDM-O9_I0,6018
-sparknlp/util.py,sha256=PDbD6c9y0IAtzXK5dzhBSqghouKHpXNVnXyWS2Y20UU,1276
+sparknlp/util.py,sha256=2Z499Psal-NuEJ4CHQNgHnAJrS73QQNyCzKPo1MavU8,2279
 sparknlp/annotator/__init__.py,sha256=G746SY8dRM_AOf-gaoSKlh7D-2TKGpqqHhGr4XF-b2A,3534
 sparknlp/annotator/chunk2_doc.py,sha256=IJ3_vQHvzjqono90AZUzZ67QSYjwquuMYbN9_HSOVcg,3141
 sparknlp/annotator/chunker.py,sha256=8nz9B7R_mxKxcfJRfKvz2x_T29W3u4izE9k0wfYPzgE,5174
+sparknlp/annotator/dataframe_optimizer.py,sha256=P4GySLzz1lRCZX0UBRF9_IDuXlRS1XvRWz-B2L0zqMA,7771
 sparknlp/annotator/date2_chunk.py,sha256=tW3m_LExmhx8LMFWOGXqMyfNRXSr2dnoEHD-6DrnpXI,3153
 sparknlp/annotator/document_character_text_splitter.py,sha256=oNrOKJAKO2h1wr0bEuSqYrrltIU_Y6J6cTHy70yKy6s,9877
 sparknlp/annotator/document_normalizer.py,sha256=hU2fG6vaPfdngQapoeSu-_zS_LiBZNp2tcVBGl6eTpk,10973
@@ -102,7 +103,7 @@ sparknlp/annotator/cv/vit_for_image_classification.py,sha256=D2V3pxAd3rBi1817lxV
 sparknlp/annotator/dependency/__init__.py,sha256=eV43oXAGaYl2N1XKIEAAZJLNP8gpHm8VxuXDeDlQzR4,774
 sparknlp/annotator/dependency/dependency_parser.py,sha256=SxyvHPp8Hs1Xnm5X1nLTMi095XoQMtfL8pbys15mYAI,11212
 sparknlp/annotator/dependency/typed_dependency_parser.py,sha256=60vPdYkbFk9MPGegg3m9Uik9cMXpMZd8tBvXG39gNww,12456
-sparknlp/annotator/embeddings/__init__.py,sha256=KHDCHb8SMlkSGGSu69SfKneUDDUlBdMGdMzDrYp_cis,2408
+sparknlp/annotator/embeddings/__init__.py,sha256=Aw1oaP5DI0OS6259c0TEZZ6j3VFSvYFEerah5a-udVw,2528
 sparknlp/annotator/embeddings/albert_embeddings.py,sha256=6Rd1LIn8oFIpq_ALcJh-RUjPEO7Ht8wsHY6JHSFyMkw,9995
 sparknlp/annotator/embeddings/auto_gguf_embeddings.py,sha256=IlqkPGOH2lmZvxEyDSGX-G90DtTFOe2Rvujfbg5zvlU,20185
 sparknlp/annotator/embeddings/bert_embeddings.py,sha256=HVUjkg56kBcpGZCo-fmPG5uatMDF3swW_lnbpy1SgSI,8463
@@ -114,9 +115,11 @@ sparknlp/annotator/embeddings/deberta_embeddings.py,sha256=_b5nzLb7heFQNN-uT2oBN
 sparknlp/annotator/embeddings/distil_bert_embeddings.py,sha256=4pyMCsbvvXYeTGIMVUir9wCDKR_1f_HKtXZrTDO1Thc,9275
 sparknlp/annotator/embeddings/doc2vec.py,sha256=Xk3MdEkXatX9lRgbFbAdnIDrLgIxzUIGWFBZeo9BTq0,13226
 sparknlp/annotator/embeddings/e5_embeddings.py,sha256=Esuvrq9JlogGaSSzFVVDkOFMwgYwFwr17I62ZiCDm0k,7858
+sparknlp/annotator/embeddings/e5v_embeddings.py,sha256=NFHO2nxDcgVzyKQ6yz1BWyqtjwt9QHwlkKbBXFwhsO8,5951
 sparknlp/annotator/embeddings/elmo_embeddings.py,sha256=KV-KPs0Pq_OpPaHsnqBz2k_S7VdzyFZ4632IeFNKqJ8,9858
 sparknlp/annotator/embeddings/instructor_embeddings.py,sha256=CTKmbuBOx_KBM4JM-Y1U5LyR-6rrnpoBGbgGE_axS1c,8670
 sparknlp/annotator/embeddings/longformer_embeddings.py,sha256=jS4fxB5O0-d9ta9VKv8ai-17n5YHt5rML8QxUw7K4Io,8754
+sparknlp/annotator/embeddings/minilm_embeddings.py,sha256=iKO3FPA6qkGjJAyPlhYpAWssnrNvlYjYxZaZEC0QDhc,7436
 sparknlp/annotator/embeddings/mpnet_embeddings.py,sha256=7d6E4lS7jjkppDPvty1UHNNrbykkriFiysrxZ_RzL0U,7875
 sparknlp/annotator/embeddings/mxbai_embeddings.py,sha256=kCaYcM3lLYJjhElLK5isdxzJqIvoGZlUKKNkySMUkE8,6017
 sparknlp/annotator/embeddings/nomic_embeddings.py,sha256=WTllH3htx9wDD2Le8pZgKVPM_U8XNmroJb6f4PeVeP8,7347
@@ -224,7 +227,7 @@ sparknlp/common/read_as.py,sha256=imxPGwV7jr4Li_acbo0OAHHRGCBbYv-akzEGaBWEfcY,12
 sparknlp/common/recursive_annotator_approach.py,sha256=vqugBw22cE3Ff7PIpRlnYFuOlchgL0nM26D8j-NdpqU,1449
 sparknlp/common/storage.py,sha256=D91H3p8EIjNspjqAYu6ephRpCUtdcAir4_PrAbkIQWE,4842
 sparknlp/common/utils.py,sha256=Yne6yYcwKxhOZC-U4qfYoDhWUP_6BIaAjI5X_P_df1E,1306
-sparknlp/internal/__init__.py,sha256=Bn-p54L1DMmrSc1nfWpmCpv_WsRZJGXLc1UmQBjMXaw,39839
+sparknlp/internal/__init__.py,sha256=wvC7ovDfII5GiYSwNpA1HHttnlXjbFgpYAGV68NsiQo,40446
 sparknlp/internal/annotator_java_ml.py,sha256=UGPoThG0rGXUOXGSQnDzEDW81Mu1s5RPF29v7DFyE3c,1187
 sparknlp/internal/annotator_transformer.py,sha256=fXmc2IWXGybqZpbEU9obmbdBYPc798y42zvSB4tqV9U,1448
 sparknlp/internal/extended_java_wrapper.py,sha256=hwP0133-hDiDf5sBF-P3MtUsuuDj1PpQbtGZQIRwzfk,2240
@@ -234,16 +237,16 @@ sparknlp/logging/__init__.py,sha256=DoROFF5KLZe4t4Q-OHxqk1nhqbw9NQ-wb64y8icNwgw,
 sparknlp/logging/comet.py,sha256=_ZBi9-hlilCAnd4lvdYMWiq4Vqsppv8kow3k0cf-NG4,15958
 sparknlp/partition/__init__.py,sha256=L0w-yv_HnnvoKlSX5MzI2GKHW3RLLfGyq8bgWYVeKjU,749
 sparknlp/partition/partition.py,sha256=GXEAUvOea04Vc_JK0z112cAKFrJ4AEpjLJ8xlzZt6Kw,8551
-sparknlp/partition/partition_properties.py,sha256=68gG_OECjij5xSuiX7O_jguFGX797U_OtOOSmT4Sdj4,7849
-sparknlp/partition/partition_transformer.py,sha256=dqvTLeBI6RNH_hGaib-AJQsWBKpj96iuvBAm_TJAuho,6774
+sparknlp/partition/partition_properties.py,sha256=xhAMhlsTBg-WS6KWDyVbRPwO7IzpowVVhJNR-ZGhvdo,9520
+sparknlp/partition/partition_transformer.py,sha256=lRR1h-IMlHR8M0VeB50SbU39GHHF5PgMaJ42qOriS6A,6855
 sparknlp/pretrained/__init__.py,sha256=GV-x9UBK8F2_IR6zYatrzFcVJtkSUIMbxqWsxRUePmQ,793
 sparknlp/pretrained/pretrained_pipeline.py,sha256=lquxiaABuA68Rmu7csamJPqBoRJqMUO0oNHsmEZDAIs,5740
 sparknlp/pretrained/resource_downloader.py,sha256=8_-rpvO2LsX_Lq4wMPif2ca3RlJZWEabt8pDm2xymiI,7806
 sparknlp/pretrained/utils.py,sha256=T1MrvW_DaWk_jcOjVLOea0NMFE9w8fe0ZT_5urZ_nEY,1099
 sparknlp/reader/__init__.py,sha256=-Toj3AIBki-zXPpV8ezFTI2LX1yP_rK2bhpoa8nBkTw,685
 sparknlp/reader/enums.py,sha256=MNGug9oJ1BBLM1Pbske13kAabalDzHa2kucF5xzFpHs,770
-sparknlp/reader/pdf_to_text.py,sha256=pI1BBQ44tXn8GIMv--_kZJ3bPP8R9Q1lYejkfhi5pMQ,5739
-sparknlp/reader/sparknlp_reader.py,sha256=ZGTpWtboPptQop-ZCF3XB3Gp21BKq6CfVBEnbb0Sb5A,15072
+sparknlp/reader/pdf_to_text.py,sha256=eWw-cwjosmcSZ9eHso0F5QQoeGBBnwsOhzhCXXvMjZA,7169
+sparknlp/reader/sparknlp_reader.py,sha256=ybnMlwJaBOVbjDw7ng39jcrshlQzexwq98_PTwVeM8g,16779
 sparknlp/training/__init__.py,sha256=qREi9u-5Vc2VjpL6-XZsyvu5jSEIdIhowW7_kKaqMqo,852
 sparknlp/training/conll.py,sha256=wKBiSTrjc6mjsl7Nyt6B8f4yXsDJkZb-sn8iOjix9cE,6961
 sparknlp/training/conllu.py,sha256=8r3i-tmyrLsyk1DtZ9uo2mMDCWb1yw2Y5W6UsV13MkY,4953
@@ -274,7 +277,7 @@ sparknlp/training/_tf_graph_builders_1x/ner_dl/dataset_encoder.py,sha256=R4yHFN3
 sparknlp/training/_tf_graph_builders_1x/ner_dl/ner_model.py,sha256=EoCSdcIjqQ3wv13MAuuWrKV8wyVBP0SbOEW41omHlR0,23189
 sparknlp/training/_tf_graph_builders_1x/ner_dl/ner_model_saver.py,sha256=k5CQ7gKV6HZbZMB8cKLUJuZxoZWlP_DFWdZ--aIDwsc,2356
 sparknlp/training/_tf_graph_builders_1x/ner_dl/sentence_grouper.py,sha256=pAxjWhjazSX8Vg0MFqJiuRVw1IbnQNSs-8Xp26L4nko,870
-spark_nlp-6.0.2.dist-info/METADATA,sha256=U7T7SvtFmoJjcaW5HjycYAS6fLQn924ok2Q4gqw4xPw,19722
-spark_nlp-6.0.2.dist-info/WHEEL,sha256=JNWh1Fm1UdwIQV075glCn4MVuCRs0sotJIq-J6rbxCU,109
-spark_nlp-6.0.2.dist-info/top_level.txt,sha256=uuytur4pyMRw2H_txNY2ZkaucZHUs22QF8-R03ch_-E,13
-spark_nlp-6.0.2.dist-info/RECORD,,
+spark_nlp-6.0.4.dist-info/METADATA,sha256=xU_AVvIsdTMbYdmEsyU-05YDMQizz-l26J6zKkoC1C8,19722
+spark_nlp-6.0.4.dist-info/WHEEL,sha256=JNWh1Fm1UdwIQV075glCn4MVuCRs0sotJIq-J6rbxCU,109
+spark_nlp-6.0.4.dist-info/top_level.txt,sha256=uuytur4pyMRw2H_txNY2ZkaucZHUs22QF8-R03ch_-E,13
+spark_nlp-6.0.4.dist-info/RECORD,,

sparknlp/__init__.py

@@ -66,7 +66,7 @@ sys.modules['com.johnsnowlabs.ml.ai'] = annotator
 annotators = annotator
 embeddings = annotator
 
-__version__ = "6.0.2"
+__version__ = "6.0.4"
 
 
 def start(gpu=False,

sparknlp/annotator/dataframe_optimizer.py

@@ -0,0 +1,216 @@
+#  Copyright 2017-2025 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+from pyspark.ml import Transformer
+from pyspark.ml.param.shared import *
+from pyspark.sql import DataFrame
+from typing import Any
+
+# Custom converter for string-to-string dictionaries
+def toStringDict(value):
+    if not isinstance(value, dict):
+        raise TypeError("Expected a dictionary of strings.")
+    return {str(k): str(v) for k, v in value.items()}
+
+class DataFrameOptimizer(Transformer):
+    """
+    Optimizes a Spark DataFrame by repartitioning, optionally caching, and persisting it to disk.
+
+    This transformer is intended to improve performance for Spark NLP pipelines or when preparing
+    data for export. It allows partition tuning via `numPartitions` directly, or indirectly using
+    `executorCores` and `numWorkers`. The DataFrame can also be persisted in a specified format
+    (`csv`, `json`, or `parquet`) with additional writer options.
+
+    Parameters
+    ----------
+    executorCores : int, optional
+        Number of cores per Spark executor (used to compute number of partitions if `numPartitions` is not set).
+
+    numWorkers : int, optional
+        Number of executor nodes (used to compute number of partitions if `numPartitions` is not set).
+
+    numPartitions : int, optional
+        Target number of partitions for the DataFrame (overrides calculation via cores × workers).
+
+    doCache : bool, default False
+        Whether to cache the DataFrame after repartitioning.
+
+    persistPath : str, optional
+        Path to save the DataFrame output (if persistence is enabled).
+
+    persistFormat : str, optional
+        Format to persist the DataFrame in: one of `'csv'`, `'json'`, or `'parquet'`.
+
+    outputOptions : dict, optional
+        Dictionary of options for the DataFrameWriter (e.g., `{"compression": "snappy"}` for parquet).
+
+    Examples
+    --------
+    >>> optimizer = DataFrameOptimizer() \\
+    ...     .setExecutorCores(4) \\
+    ...     .setNumWorkers(5) \\
+    ...     .setDoCache(True) \\
+    ...     .setPersistPath("/tmp/out") \\
+    ...     .setPersistFormat("parquet") \\
+    ...     .setOutputOptions({"compression": "snappy"})
+
+    >>> optimized_df = optimizer.transform(input_df)
+
+    Notes
+    -----
+    - You must specify either `numPartitions`, or both `executorCores` and `numWorkers`.
+    - Schema is preserved; no columns are modified or removed.
+    """
+
+    executorCores = Param(
+        Params._dummy(),
+        "executorCores",
+        "Number of cores per executor",
+        typeConverter = TypeConverters.toInt
+    )
+    numWorkers = Param(
+        Params._dummy(),
+        "numWorkers",
+        "Number of Spark workers",
+        typeConverter = TypeConverters.toInt
+    )
+    numPartitions = Param(
+        Params._dummy(),
+        "numPartitions",
+        "Total number of partitions (overrides executorCores * numWorkers)",
+        typeConverter = TypeConverters.toInt
+    )
+    doCache = Param(
+        Params._dummy(),
+        "doCache",
+        "Whether to cache the DataFrame",
+        typeConverter = TypeConverters.toBoolean
+    )
+
+    persistPath = Param(
+        Params._dummy(),
+        "persistPath",
+        "Optional path to persist the DataFrame",
+        typeConverter = TypeConverters.toString
+    )
+    persistFormat = Param(
+        Params._dummy(),
+        "persistFormat",
+        "Format to persist: parquet, json, csv",
+        typeConverter = TypeConverters.toString
+    )
+
+    outputOptions = Param(
+        Params._dummy(),
+        "outputOptions",
+        "Additional writer options",
+        typeConverter=toStringDict
+    )
+
+    def __init__(self):
+        super().__init__()
+        self._setDefault(
+            doCache=False,
+            persistFormat="none",
+            numPartitions=1,
+            executorCores=1,
+            numWorkers=1
+        )
+
+    # Parameter setters
+    def setExecutorCores(self, value: int):
+        """Set the number of executor cores."""
+        return self._set(executorCores=value)
+
+    def setNumWorkers(self, value: int):
+        """Set the number of Spark workers."""
+        return self._set(numWorkers=value)
+
+    def setNumPartitions(self, value: int):
+        """Set the total number of partitions (overrides cores * workers)."""
+        return self._set(numPartitions=value)
+
+    def setDoCache(self, value: bool):
+        """Set whether to cache the DataFrame."""
+        return self._set(doCache=value)
+
+    def setPersistPath(self, value: str):
+        """Set the path where the DataFrame should be persisted."""
+        return self._set(persistPath=value)
+
+    def setPersistFormat(self, value: str):
+        """Set the format to persist the DataFrame (parquet, json, csv)."""
+        return self._set(persistFormat=value)
+
+    def setOutputOptions(self, value: dict):
+        """Set additional writer options (e.g. for csv headers)."""
+        return self._set(outputOptions=value)
+
+    # Optional bulk setter
+    def setParams(self, **kwargs: Any):
+        for param, value in kwargs.items():
+            self._set(**{param: value})
+        return self
+
+    def _transform(self, dataset: DataFrame) -> DataFrame:
+        self._validate_params()
+        part_count = self.getOrDefault(self.numPartitions)
+        cores = self.getOrDefault(self.executorCores)
+        workers = self.getOrDefault(self.numWorkers)
+        if cores is None or workers is None:
+            raise ValueError("Provide either numPartitions or both executorCores and numWorkers")
+        if part_count == 1:
+            part_count = cores * workers
+
+        optimized_df = dataset.repartition(part_count)
+
+        if self.getOrDefault(self.doCache):
+            optimized_df = optimized_df.cache()
+
+        format = self.getOrDefault(self.persistFormat).lower()
+        if format != "none":
+            path = self.getOrDefault(self.persistPath)
+            if not path:
+                raise ValueError("persistPath must be set when persistFormat is not 'none'")
+            writer = optimized_df.write.mode("overwrite")
+            if self.isDefined(self.outputOptions):
+                writer = writer.options(**self.getOrDefault(self.outputOptions))
+            if format == "parquet":
+                writer.parquet(path)
+            elif format == "json":
+                writer.json(path)
+            elif format == "csv":
+                writer.csv(path)
+            else:
+                raise ValueError(f"Unsupported format: {format}")
+
+        return optimized_df
+
+    def _validate_params(self):
+        if self.isDefined(self.executorCores):
+            val = self.getOrDefault(self.executorCores)
+            if val <= 0:
+                raise ValueError("executorCores must be > 0")
+
+        if self.isDefined(self.numWorkers):
+            val = self.getOrDefault(self.numWorkers)
+            if val <= 0:
+                raise ValueError("numWorkers must be > 0")
+
+        if self.isDefined(self.numPartitions):
+            val = self.getOrDefault(self.numPartitions)
+            if val <= 0:
+                raise ValueError("numPartitions must be > 0")
+
+        if self.isDefined(self.persistPath) and not self.isDefined(self.persistFormat):
+            raise ValueError("persistFormat must be defined when persistPath is set")

sparknlp/annotator/embeddings/__init__.py

@@ -25,6 +25,7 @@ from sparknlp.annotator.embeddings.elmo_embeddings import *
 from sparknlp.annotator.embeddings.e5_embeddings import *
 from sparknlp.annotator.embeddings.instructor_embeddings import *
 from sparknlp.annotator.embeddings.longformer_embeddings import *
+from sparknlp.annotator.embeddings.minilm_embeddings import *
 from sparknlp.annotator.embeddings.mpnet_embeddings import *
 from sparknlp.annotator.embeddings.roberta_embeddings import *
 from sparknlp.annotator.embeddings.roberta_sentence_embeddings import *
@@ -41,3 +42,4 @@ from sparknlp.annotator.embeddings.mxbai_embeddings import *
 from sparknlp.annotator.embeddings.snowflake_embeddings import *
 from sparknlp.annotator.embeddings.nomic_embeddings import *
 from sparknlp.annotator.embeddings.auto_gguf_embeddings import *
+from sparknlp.annotator.embeddings.e5v_embeddings import *

sparknlp/annotator/embeddings/e5v_embeddings.py

@@ -0,0 +1,138 @@
+#  Copyright 2017-2024 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+from sparknlp.common import *
+
+class E5VEmbeddings(AnnotatorModel,
+                   HasBatchedAnnotateImage,
+                   HasImageFeatureProperties,
+                   HasEngine,
+                    HasRescaleFactor):
+    """Universal multimodal embeddings using the E5-V model (see https://huggingface.co/royokong/e5-v).
+
+    E5-V bridges the modality gap between different input types (text, image) and demonstrates strong performance in multimodal embeddings, even without fine-tuning. It also supports a single-modality training approach, where the model is trained exclusively on text pairs, often yielding better performance than multimodal training.
+
+    Pretrained models can be loaded with :meth:`.pretrained` of the companion object:
+
+    >>> e5vEmbeddings = E5VEmbeddings.pretrained() \
+    ...     .setInputCols(["image_assembler"]) \
+    ...     .setOutputCol("e5v")
+
+    The default model is ``"e5v_int4"``, if no name is provided.
+
+    For available pretrained models please see the `Models Hub <https://sparknlp.org/models?task=Question+Answering>`__.
+
+    ====================== ======================
+    Input Annotation types Output Annotation type
+    ====================== ======================
+    ``IMAGE``              ``SENTENCE_EMBEDDINGS``
+    ====================== ======================
+
+    Examples
+    --------
+    Image + Text Embedding:
+    >>> import sparknlp
+    >>> from sparknlp.base import *
+    >>> from sparknlp.annotator import *
+    >>> from pyspark.ml import Pipeline
+    >>> image_df = spark.read.format("image").option("dropInvalid", value = True).load(imageFolder)
+    >>> imagePrompt = "<|start_header_id|>user<|end_header_id|>\n\n<image>\\nSummary above image in one word: <|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n \n"
+    >>> test_df = image_df.withColumn("text", lit(imagePrompt))
+    >>> imageAssembler = ImageAssembler() \
+    ...     .setInputCol("image") \
+    ...     .setOutputCol("image_assembler")
+    >>> e5vEmbeddings = E5VEmbeddings.pretrained() \
+    ...     .setInputCols(["image_assembler"]) \
+    ...     .setOutputCol("e5v")
+    >>> pipeline = Pipeline().setStages([
+    ...     imageAssembler,
+    ...     e5vEmbeddings
+    ... ])
+    >>> result = pipeline.fit(test_df).transform(test_df)
+    >>> result.select("e5v.embeddings").show(truncate = False)
+
+    Text-Only Embedding:
+    >>> from sparknlp.util import EmbeddingsDataFrameUtils
+    >>> textPrompt = "<|start_header_id|>user<|end_header_id|>\n\n<sent>\\nSummary above sentence in one word: <|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n \n"
+    >>> textDesc = "A cat sitting in a box."
+    >>> nullImageDF = spark.createDataFrame(spark.sparkContext.parallelize([EmbeddingsDataFrameUtils.emptyImageRow]), EmbeddingsDataFrameUtils.imageSchema)
+    >>> textDF = nullImageDF.withColumn("text", lit(textPrompt.replace("<sent>", textDesc)))
+    >>> e5vEmbeddings = E5VEmbeddings.pretrained() \
+    ...     .setInputCols(["image"]) \
+    ...     .setOutputCol("e5v")
+    >>> result = e5vEmbeddings.transform(textDF)
+    >>> result.select("e5v.embeddings").show(truncate = False)
+    """
+
+    name = "E5VEmbeddings"
+
+    inputAnnotatorTypes = [AnnotatorType.IMAGE]
+    outputAnnotatorType = AnnotatorType.SENTENCE_EMBEDDINGS
+
+    @keyword_only
+    def __init__(self, classname="com.johnsnowlabs.nlp.embeddings.E5VEmbeddings", java_model=None):
+        """Initializes the E5VEmbeddings annotator.
+
+        Parameters
+        ----------
+        classname : str, optional
+            The Java class name of the annotator, by default "com.johnsnowlabs.nlp.annotators.embeddings.E5VEmbeddings"
+        java_model : Optional[java.lang.Object], optional
+            A pre-initialized Java model, by default None
+        """
+        super(E5VEmbeddings, self).__init__(classname=classname, java_model=java_model)
+        self._setDefault()
+
+    @staticmethod
+    def loadSavedModel(folder, spark_session, use_openvino=False):
+        """Loads a locally saved model.
+
+        Parameters
+        ----------
+        folder : str
+            Folder of the saved model
+        spark_session : pyspark.sql.SparkSession
+            The current SparkSession
+        use_openvino : bool, optional
+            Whether to use OpenVINO engine, by default False
+
+        Returns
+        -------
+        E5VEmbeddings
+            The restored model
+        """
+        from sparknlp.internal import _E5VEmbeddingsLoader
+        jModel = _E5VEmbeddingsLoader(folder, spark_session._jsparkSession, use_openvino)._java_obj
+        return E5VEmbeddings(java_model=jModel)
+
+    @staticmethod
+    def pretrained(name="e5v_int4", lang="en", remote_loc=None):
+        """Downloads and loads a pretrained model.
+
+        Parameters
+        ----------
+        name : str, optional
+            Name of the pretrained model, by default "e5v_int4"
+        lang : str, optional
+            Language of the pretrained model, by default "en"
+        remote_loc : str, optional
+            Optional remote address of the resource, by default None. Will use Spark NLPs repositories otherwise.
+
+        Returns
+        -------
+        E5VEmbeddings
+            The restored model
+        """
+        from sparknlp.pretrained import ResourceDownloader
+        return ResourceDownloader.downloadModel(E5VEmbeddings, name, lang, remote_loc) 

sparknlp/annotator/embeddings/minilm_embeddings.py

@@ -0,0 +1,189 @@
+#  Copyright 2017-2022 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+"""Contains classes for MiniLMEmbeddings."""
+
+from sparknlp.common import *
+
+
+class MiniLMEmbeddings(AnnotatorModel,
+                           HasEmbeddingsProperties,
+                           HasCaseSensitiveProperties,
+                           HasStorageRef,
+                           HasBatchedAnnotate,
+                           HasMaxSentenceLengthLimit):
+    """Sentence embeddings using MiniLM.
+
+    MiniLM, a lightweight and efficient sentence embedding model that can generate text embeddings for various NLP tasks (e.g., classification, retrieval, clustering, text evaluation, etc.)
+    Note that this annotator is only supported for Spark Versions 3.4 and up.
+
+    Pretrained models can be loaded with :meth:`.pretrained` of the companion
+    object:
+
+    >>> embeddings = MiniLMEmbeddings.pretrained() \\
+    ...     .setInputCols(["document"]) \\
+    ...     .setOutputCol("minilm_embeddings")
+
+
+    The default model is ``"minilm_l6_v2"``, if no name is provided.
+
+    For available pretrained models please see the
+    `Models Hub <https://sparknlp.org/models?q=MiniLM>`__.
+
+
+    ====================== ======================
+    Input Annotation types Output Annotation type
+    ====================== ======================
+    ``DOCUMENT``            ``SENTENCE_EMBEDDINGS``
+    ====================== ======================
+
+    Parameters
+    ----------
+    batchSize
+        Size of every batch , by default 8
+    dimension
+        Number of embedding dimensions, by default 384
+    caseSensitive
+        Whether to ignore case in tokens for embeddings matching, by default False
+    maxSentenceLength
+        Max sentence length to process, by default 512
+    configProtoBytes
+        ConfigProto from tensorflow, serialized into byte array.
+
+    References
+    ----------
+    `MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers <https://arxiv.org/abs/2002.10957>`__
+
+    `MiniLM Github Repository <https://github.com/microsoft/unilm/tree/master/minilm>`__
+
+    **Paper abstract**
+
+    *We present a simple and effective approach to compress large pre-trained Transformer models
+    by distilling the self-attention module of the last Transformer layer. The compressed model
+    (called MiniLM) can be trained with task-agnostic distillation and then fine-tuned on various
+    downstream tasks. We evaluate MiniLM on the GLUE benchmark and show that it achieves comparable
+    results with BERT-base while being 4.3x smaller and 5.5x faster. We also show that MiniLM can
+    be further compressed to 22x smaller and 12x faster than BERT-base while maintaining comparable
+    performance.*
+
+    Examples
+    --------
+    >>> import sparknlp
+    >>> from sparknlp.base import *
+    >>> from sparknlp.annotator import *
+    >>> from pyspark.ml import Pipeline
+    >>> documentAssembler = DocumentAssembler() \\
+    ...     .setInputCol("text") \\
+    ...     .setOutputCol("document")
+    >>> embeddings = MiniLMEmbeddings.pretrained() \\
+    ...     .setInputCols(["document"]) \\
+    ...     .setOutputCol("minilm_embeddings")
+    >>> embeddingsFinisher = EmbeddingsFinisher() \\
+    ...     .setInputCols(["minilm_embeddings"]) \\
+    ...     .setOutputCols("finished_embeddings") \\
+    ...     .setOutputAsVector(True)
+    >>> pipeline = Pipeline().setStages([
+    ...     documentAssembler,
+    ...     embeddings,
+    ...     embeddingsFinisher
+    ... ])
+    >>> data = spark.createDataFrame([["This is a sample sentence for embedding generation.",
+    ... "Another example sentence to demonstrate MiniLM embeddings.",
+    ... ]]).toDF("text")
+    >>> result = pipeline.fit(data).transform(data)
+    >>> result.selectExpr("explode(finished_embeddings) as result").show(5, 80)
+    +--------------------------------------------------------------------------------+
+    |                                                                          result|
+    +--------------------------------------------------------------------------------+
+    |[[0.1234567, -0.2345678, 0.3456789, -0.4567890, 0.5678901, -0.6789012...|
+    |[[0.2345678, -0.3456789, 0.4567890, -0.5678901, 0.6789012, -0.7890123...|
+    +--------------------------------------------------------------------------------+
+    """
+
+    name = "MiniLMEmbeddings"
+
+    inputAnnotatorTypes = [AnnotatorType.DOCUMENT]
+
+    outputAnnotatorType = AnnotatorType.SENTENCE_EMBEDDINGS
+    configProtoBytes = Param(Params._dummy(),
+                             "configProtoBytes",
+                             "ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()",
+                             TypeConverters.toListInt)
+
+
+    def setConfigProtoBytes(self, b):
+        """Sets configProto from tensorflow, serialized into byte array.
+
+        Parameters
+        ----------
+        b : List[int]
+            ConfigProto from tensorflow, serialized into byte array
+        """
+        return self._set(configProtoBytes=b)
+
+    @keyword_only
+    def __init__(self, classname="com.johnsnowlabs.nlp.embeddings.MiniLMEmbeddings", java_model=None):
+        super(MiniLMEmbeddings, self).__init__(
+            classname=classname,
+            java_model=java_model
+        )
+        self._setDefault(
+            dimension=384,
+            batchSize=8,
+            maxSentenceLength=512,
+            caseSensitive=False,
+        )
+
+    @staticmethod
+    def loadSavedModel(folder, spark_session, use_openvino=False):
+        """Loads a locally saved model.
+
+        Parameters
+        ----------
+        folder : str
+            Folder of the saved model
+        spark_session : pyspark.sql.SparkSession
+            The current SparkSession
+        use_openvino : bool
+            Use OpenVINO backend
+
+        Returns
+        -------
+        MiniLMEmbeddings
+            The restored model
+        """
+        from sparknlp.internal import _MiniLMLoader
+        jModel = _MiniLMLoader(folder, spark_session._jsparkSession, use_openvino)._java_obj
+        return MiniLMEmbeddings(java_model=jModel)
+
+    @staticmethod
+    def pretrained(name="minilm_l6_v2", lang="en", remote_loc=None):
+        """Downloads and loads a pretrained model.
+
+        Parameters
+        ----------
+        name : str, optional
+            Name of the pretrained model, by default "minilm_l6_v2"
+        lang : str, optional
+            Language of the pretrained model, by default "en"
+        remote_loc : str, optional
+            Optional remote address of the resource, by default None. Will use
+            Spark NLPs repositories otherwise.
+
+        Returns
+        -------
+        MiniLMEmbeddings
+            The restored model
+        """
+        from sparknlp.pretrained import ResourceDownloader
+        return ResourceDownloader.downloadModel(MiniLMEmbeddings, name, lang, remote_loc)

sparknlp/internal/__init__.py

@@ -257,6 +257,16 @@ class _E5Loader(ExtendedJavaWrapper):
         )
 
 
+class _MiniLMLoader(ExtendedJavaWrapper):
+    def __init__(self, path, jspark, use_openvino=False):
+        super(_MiniLMLoader, self).__init__(
+            "com.johnsnowlabs.nlp.embeddings.MiniLMEmbeddings.loadSavedModel",
+            path,
+            jspark,
+            use_openvino,
+        )
+
+
 class _BGELoader(ExtendedJavaWrapper):
     def __init__(self, path, jspark):
         super(_BGELoader, self).__init__(
@@ -1165,3 +1175,11 @@ class _Florence2TransformerLoader(ExtendedJavaWrapper):
             jspark,
             use_openvino,
         )
+class _E5VEmbeddingsLoader(ExtendedJavaWrapper):
+    def __init__(self, path, jspark, use_openvino=False):
+        super(_E5VEmbeddingsLoader, self).__init__(
+            "com.johnsnowlabs.nlp.embeddings.E5VEmbeddings.loadSavedModel",
+            path,
+            jspark,
+            use_openvino
+        )

sparknlp/partition/partition_properties.py

@@ -254,4 +254,66 @@ class HasTextReaderProperties(Params):
         return self._set(threshold=value)
 
     def getThreshold(self):
-        return self.getOrDefault(self.threshold)
+        return self.getOrDefault(self.threshold)
+
+class HasChunkerProperties(Params):
+
+    chunkingStrategy = Param(
+        Params._dummy(),
+        "chunkingStrategy",
+        "Set the chunking strategy",
+        typeConverter=TypeConverters.toString
+    )
+
+    def setChunkingStrategy(self, value):
+        return self._set(chunkingStrategy=value)
+
+    maxCharacters = Param(
+        Params._dummy(),
+        "maxCharacters",
+        "Set the maximum number of characters",
+        typeConverter=TypeConverters.toInt
+    )
+
+    def setMaxCharacters(self, value):
+        return self._set(maxCharacters=value)
+
+    newAfterNChars = Param(
+        Params._dummy(),
+        "newAfterNChars",
+        "Insert a new chunk after N characters",
+        typeConverter=TypeConverters.toInt
+    )
+
+    def setNewAfterNChars(self, value):
+        return self._set(newAfterNChars=value)
+
+    overlap = Param(
+        Params._dummy(),
+        "overlap",
+        "Set the number of overlapping characters between chunks",
+        typeConverter=TypeConverters.toInt
+    )
+
+    def setOverlap(self, value):
+        return self._set(overlap=value)
+
+    combineTextUnderNChars = Param(
+        Params._dummy(),
+        "combineTextUnderNChars",
+        "Threshold to merge adjacent small sections",
+        typeConverter=TypeConverters.toInt
+    )
+
+    def setCombineTextUnderNChars(self, value):
+        return self._set(combineTextUnderNChars=value)
+
+    overlapAll = Param(
+        Params._dummy(),
+        "overlapAll",
+        "Apply overlap context between all sections, not just split chunks",
+        typeConverter=TypeConverters.toBoolean
+    )
+
+    def setOverlapAll(self, value):
+        return self._set(overlapAll=value)

sparknlp/partition/partition_transformer.py

@@ -15,13 +15,15 @@
 from sparknlp.common import *
 from sparknlp.partition.partition_properties import *
 
+
 class PartitionTransformer(
     AnnotatorModel,
     HasEmailReaderProperties,
     HasExcelReaderProperties,
     HasHTMLReaderProperties,
     HasPowerPointProperties,
-    HasTextReaderProperties
+    HasTextReaderProperties,
+    HasChunkerProperties
 ):
     """
     The PartitionTransformer annotator allows you to use the Partition feature more smoothly
@@ -162,10 +164,6 @@ class PartitionTransformer(
     def getIncludePageBreaks(self):
         return self.getOrDefault(self.includePageBreaks)
 
-    # def setHeaders(self, headers: Dict[str, str]):
-    #     self._call_java("setHeadersPython", headers)
-    #     return self
-
     @keyword_only
     def __init__(self, classname="com.johnsnowlabs.partition.PartitionTransformer",
                  java_model=None):
@@ -192,5 +190,11 @@ class PartitionTransformer(
             paragraphSplit=DOUBLE_PARAGRAPH_PATTERN,
             shortLineWordThreshold=5,
             maxLineCount=2000,
-            threshold=0.1
-        )
+            threshold=0.1,
+            chunkingStrategy="",
+            maxCharacters=100,
+            newAfterNChars=-1,
+            overlap=0,
+            combineTextUnderNChars=0,
+            overlapAll=False
+        )

sparknlp/reader/pdf_to_text.py

@@ -1,3 +1,16 @@
+#  Copyright 2017-2025 John Snow Labs
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
 from pyspark import keyword_only
 from pyspark.ml.param import Param, Params, TypeConverters
 from pyspark.ml.param.shared import HasInputCol, HasOutputCol
@@ -89,6 +102,14 @@ class PdfToText(JavaTransformer, HasInputCol, HasOutputCol,
                         "Force to extract only number of pages",
                         typeConverter=TypeConverters.toBoolean)
 
+    extractCoordinates = Param(Params._dummy(), "extractCoordinates",
+                               "Force extract coordinates of text.",
+                               typeConverter=TypeConverters.toBoolean)
+
+    normalizeLigatures = Param(Params._dummy(), "normalizeLigatures",
+                               "Whether to convert ligature chars such as 'fl' into its corresponding chars (e.g., {'f', 'l'}).",
+                               typeConverter=TypeConverters.toBoolean)
+
     @keyword_only
     def __init__(self):
         """
@@ -154,4 +175,16 @@ class PdfToText(JavaTransformer, HasInputCol, HasOutputCol,
         """
         Sets the value of :py:attr:`sort`.
         """
-        return self._set(sort=value)
+        return self._set(sort=value)
+
+    def setExtractCoordinates(self, value):
+        """
+        Sets the value of :py:attr:`extractCoordinates`.
+        """
+        return self._set(extractCoordinates=value)
+
+    def setNormalizeLigatures(self, value):
+        """
+        Sets the value of :py:attr:`normalizeLigatures`.
+        """
+        return self._set(normalizeLigatures=value)

sparknlp/reader/sparknlp_reader.py

@@ -322,4 +322,49 @@ class SparkNLPReader(ExtendedJavaWrapper):
         if not isinstance(docPath, str):
             raise TypeError("docPath must be a string")
         jdf = self._java_obj.txt(docPath)
+        return self.getDataFrame(self.spark, jdf)
+
+    def xml(self, docPath):
+        """Reads XML files and returns a Spark DataFrame.
+
+        Parameters
+        ----------
+        docPath : str
+            Path to an XML file or a directory containing XML files.
+
+        Returns
+        -------
+        pyspark.sql.DataFrame
+            A DataFrame containing parsed XML content.
+
+        Examples
+        --------
+        >>> from sparknlp.reader import SparkNLPReader
+        >>> xml_df = SparkNLPReader(spark).xml("home/user/xml-directory")
+
+        You can use SparkNLP for one line of code
+
+        >>> import sparknlp
+        >>> xml_df = sparknlp.read().xml("home/user/xml-directory")
+        >>> xml_df.show(truncate=False)
+        +-----------------------------------------------------------+
+        |xml                                                       |
+        +-----------------------------------------------------------+
+        |[{Title, John Smith, {elementId -> ..., tag -> title}}]   |
+        +-----------------------------------------------------------+
+
+        >>> xml_df.printSchema()
+        root
+         |-- path: string (nullable = true)
+         |-- xml: array (nullable = true)
+         |    |-- element: struct (containsNull = true)
+         |    |    |-- elementType: string (nullable = true)
+         |    |    |-- content: string (nullable = true)
+         |    |    |-- metadata: map (nullable = true)
+         |    |    |    |-- key: string
+         |    |    |    |-- value: string (valueContainsNull = true)
+        """
+        if not isinstance(docPath, str):
+            raise TypeError("docPath must be a string")
+        jdf = self._java_obj.xml(docPath)
         return self.getDataFrame(self.spark, jdf)

sparknlp/util.py

@@ -15,6 +15,9 @@
 
 
 import sparknlp.internal as _internal
+import numpy as np
+from pyspark.sql import Row
+from pyspark.sql.types import StructType, StructField, StringType, IntegerType, BinaryType
 
 
 def get_config_path():
@@ -33,3 +36,26 @@ class CoNLLGenerator:
             _internal._CoNLLGeneratorExportFromTargetAndPipeline(*args).apply()
         else:
             raise NotImplementedError(f"No exportConllFiles alternative takes {num_args} parameters")
+
+
+class EmbeddingsDataFrameUtils:
+    """
+    Utility for creating DataFrames compatible with multimodal embedding models (e.g., E5VEmbeddings) for text-only scenarios.
+    Provides:
+      - imageSchema: the expected schema for Spark image DataFrames
+      - emptyImageRow: a dummy image row for text-only embedding
+    """
+    imageSchema = StructType([
+        StructField(
+            "image",
+            StructType([
+                StructField("origin", StringType(), True),
+                StructField("height", IntegerType(), True),
+                StructField("width", IntegerType(), True),
+                StructField("nChannels", IntegerType(), True),
+                StructField("mode", IntegerType(), True),
+                StructField("data", BinaryType(), True),
+            ]),
+        )
+    ])
+    emptyImageRow = Row(Row("", 0, 0, 0, 0, bytes()))