py2ls 0.2.4.24__py3-none-any.whl → 0.2.4.25__py3-none-any.whl

py2ls/nl2ls.py

@@ -0,0 +1,283 @@
+from . import translator,ips,plot
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt 
+
+
+def detect(text, method: str = "vader", nb_model=None, device=-1,overall_method="major",overall_threhold=0.8,overall_weight=None,plot_=True,verbose=True, **kwargs) -> dict:
+    """
+    Analyze the sentiment of a text or a list of texts using different methods.
+
+    Parameters:
+    - text (str or list of str): The text(s) to analyze. Can be a single text or a list of texts.
+    - method (str): The method to use ('vader', 'textblob', 'naive_bayes', 'transformers', 'senta').
+    - nb_model (Optional[MultinomialNB]): Pre-trained Naive Bayes model (required if method='naive_bayes').
+    - vectorizer (Optional[TfidfVectorizer]): Vectorizer trained with Naive Bayes model (required if method='naive_bayes').
+    - device (int): Device to run the model on (-1 for CPU, 0 for GPU).
+    - transformer_model_name (str): Transformer model name for 'transformers' method.
+
+    Returns:
+    - dict: A dictionary with sentiment score, sentiment label, analysis method, and language.
+    """
+    result = {
+        "method": method,
+        "score": None,
+        "label": None,
+        "language": None,
+    }
+
+    methods=['vader','textblob','naive_bayes','transformer(not ready)','senta(not ready)']
+    if ips.run_once_within(10, reverse=True) and verbose:
+        print(f"methods: {methods}")
+        
+    overall_methods=["majority","average","mean","threshold","weighted","detailed"]
+    if ips.run_once_within(10, reverse=True) and verbose:
+        print(f"overall_methods: {overall_methods}")
+    # If the input is a list of texts, loop through each one
+    if isinstance(text, list):
+        results = []
+        for text_ in text:
+            results.append(detect_single_text(text_, method=method, nb_model=nb_model, device=device, **kwargs))
+        res_overall=get_overall_results(results, method=overall_method, threshold=overall_threhold, weight=overall_weight)
+        if plot_:
+            res_detail=get_overall_results(results, method='detail', threshold=overall_threhold, weight=overall_weight)
+            plot.pie(res_detail["label"].value_counts(),explode=None,verbose=False)
+        return res_overall
+    else:
+        return detect_single_text(text=text, method=method, nb_model=nb_model, device=device, **kwargs)
+
+
+def detect_single_text(text: str, method: str = "vader", nb_model=None, device=-1, **kwargs) -> dict:
+    """
+    Analyze the sentiment of a text using different methods.
+
+    Parameters:
+    - text (str): The text to analyze.
+    - method (str): The method to use ('vader', 'textblob', 'naive_bayes', 'transformers').
+    - nb_model (Optional[MultinomialNB]): Pre-trained Naive Bayes model (required if method='naive_bayes').
+    - vectorizer (Optional[TfidfVectorizer]): Vectorizer trained with Naive Bayes model (required if method='naive_bayes').
+    - transformer_model_name (str): Transformer model name for 'transformers' method.
+
+    Returns:
+    - dict: A dictionary with sentiment score, sentiment label, analysis method, and language.
+    """
+    result = {
+        "text":text,
+        "method": method,
+        "score": None,
+        "label": None,
+        "language": None,
+    }
+
+    # Detect language for additional insights
+    language = translator.detect_lang(text)
+    result["language"] = language
+    if language != "English" and method in ["vader", "textblob", "naive_bayes"]:
+        print("Detected non-English language, results may be inaccurate.")
+    methods=['vader','textblob','naive_bayes','transformer(not ready)','senta(not ready)'] 
+    method=ips.strcmp(method,methods)[0]
+    if method == "vader":
+        import nltk, os
+        from nltk.sentiment import SentimentIntensityAnalyzer
+
+        # check if it is downloaded
+        is_local = os.path.isfile(
+            os.path.join(nltk.data.path[0], "sentiment", "vader_lexicon.zip")
+        )
+        if not is_local:
+            nltk.download("vader_lexicon")
+        try:
+            sia = SentimentIntensityAnalyzer()
+            scores = sia.polarity_scores(text)
+            result["score"] = scores["compound"]
+            result["label"] = (
+                "Positive"
+                if scores["compound"] >= 0.05
+                else "Negative" if scores["compound"] <= -0.05 else "Neutral"
+            )
+        except Exception as e:
+            print(f"Error in VADER analysis: {e}")
+
+    elif method == "textblob":
+        from textblob import TextBlob
+
+        try:
+            blob = TextBlob(text)
+            polarity = blob.sentiment.polarity
+            result["score"] = polarity
+            result["label"] = (
+                "Positive"
+                if polarity > 0
+                else "Negative" if polarity < 0 else "Neutral"
+            )
+        except Exception as e:
+            print(f"Error in TextBlob analysis: {e}")
+
+    elif method == "naive_bayes":
+        from sklearn.naive_bayes import MultinomialNB
+        from sklearn.feature_extraction.text import TfidfVectorizer
+
+        try:
+            if nb_model is None or vectorizer is None:
+                from sklearn.model_selection import train_test_split
+
+                # Sample data for Naive Bayes training if model not provided
+                sample_texts = [
+                    "I love this product",
+                    "I hate this product",
+                    "It's okay, not great",
+                    "Absolutely fantastic!",
+                    "Not satisfied",
+                ]
+                sample_labels = [1, 0, 0, 1, 0]  # 1 = Positive, 0 = Negative
+
+                # Train Naive Bayes model
+                vectorizer = TfidfVectorizer()
+                X_train_tfidf = vectorizer.fit_transform(sample_texts)
+                nb_model = MultinomialNB()
+                nb_model.fit(X_train_tfidf, sample_labels)
+
+            transformed_text = vectorizer.transform([text])
+            prediction = nb_model.predict(transformed_text)[0]
+            result["score"] = max(nb_model.predict_proba(transformed_text)[0])
+            result["label"] = "Positive" if prediction == 1 else "Negative"
+
+        except Exception as e:
+            print(f"Error in Naive Bayes analysis: {e}")
+    elif method=="transformer":
+        try:
+            from transformers import pipeline
+            # Load pre-trained sentiment analysis pipeline with a Chinese model
+            classifier = pipeline('sentiment-analysis', model='bert-base-chinese', device=device)
+            analysis_result = classifier(text)
+            result["score"] = analysis_result[0]['score']
+            result["label"] = analysis_result[0]['label']
+        except Exception as e:
+            print(f"Error in Transformer analysis: {e}")
+    elif method == "senta":
+        from transformers import pipeline
+
+        try:
+            # Load the Senta model for sentiment analysis
+            classifier = pipeline('sentiment-analysis', model='junnyu/senta', device=device)
+            analysis_result = classifier(text)
+            
+            # Senta model output will be a list with one result (since it's single text input)
+            result["score"] = analysis_result[0]["score"]
+            result["label"] = analysis_result[0]["label"]
+            
+        except Exception as e:
+            print(f"Error in Senta analysis: {e}")
+
+    else:
+        print(
+            f"Unknown method '{method}'. Available methods: 'vader', 'textblob', 'naive_bayes', 'transformers'"
+        )
+        raise ValueError(
+            f"Unknown method '{method}'. Available methods: 'vader', 'textblob', 'naive_bayes', 'transformers'"
+        )
+
+    return result
+
+def get_overall_results(results, method="majority", threshold=0.8, weight=None,verbose=False):
+    from collections import Counter
+    """
+    Aggregates sentiment analysis results based on the selected method.
+
+    Parameters:
+    - results (list): A list of sentiment analysis results, each being a dictionary.
+    - method (str): The aggregation method to use ('majority', 'average', 'threshold', 'weighted', 'detailed').
+    - threshold (float): Confidence threshold for 'threshold' method.
+    - weight (dict): Optional dictionary for weighted aggregation (e.g., model name as key and weight as value).
+
+    Returns:
+    - dict: Aggregated sentiment result with final label and score.
+    """
+    def majority_voting(results):
+        """Aggregates sentiment using majority voting."""
+        labels = [result['label'] for result in results]
+        label_counts = Counter(labels)
+        final_label = label_counts.most_common(1)[0][0]  # Get the most common label
+        return {"label": final_label}
+
+
+    def average_score(results):
+        """Aggregates sentiment by calculating the average score."""
+        scores = [result['score'] for result in results]
+        avg_score = sum(scores) / len(scores)
+        
+        if avg_score > 0.05:
+            label = 'Positive'
+        elif avg_score < -0.05:
+            label = 'Negative'
+        else:
+            label = 'Neutral'
+            
+        return {"score": avg_score, "label": label}
+
+
+    def confidence_threshold(results, threshold=0.8):
+        """Aggregates sentiment based on a confidence threshold."""
+        labels = [result['label'] for result in results]
+        label_counts = Counter(labels)
+        total_results = len(results)
+        
+        for label, count in label_counts.items():
+            if count / total_results >= threshold:
+                return {"label": label}
+        
+        return {"label": 'Neutral'}  # If no label exceeds the threshold, return neutral
+
+
+    def weighted_average(results, weight=None):
+        """Aggregates sentiment based on a weighted average."""
+        if weight is None:
+            weight = {"vader": 2} 
+
+        weighted_scores = 0
+        total_weight = 0
+        
+        for result in results:
+            model = result.get('method', 'default')
+            model_weight = weight.get(model, 1)  # Default weight is 1 if model not in weight dict
+            weighted_scores += result['score'] * model_weight
+            total_weight += model_weight
+        
+        avg_weighted_score = weighted_scores / total_weight
+        
+        # Assign label based on weighted average score
+        if avg_weighted_score > 0.05:
+            label = 'Positive'
+        elif avg_weighted_score < -0.05:
+            label = 'Negative'
+        else:
+            label = 'Neutral'
+        
+        return {"score": avg_weighted_score, "label": label}
+
+    def detailed_output(results,verbose=False):
+        """Prints the detailed sentiment results."""
+        for result in results:
+            if verbose:
+                print(f"Label: {result['label']} | Score: {result['score']}")
+        return {"detailed_results": results}
+    overall_methods=["majority","average","mean","threshold","weighted","detailed"] 
+    method=ips.strcmp(method, overall_methods)[0]
+    if method == "majority":
+        return majority_voting(results)
+
+    elif method in ["mean","average"]:
+        return average_score(results)
+
+    elif method == "threshold":
+        return confidence_threshold(results, threshold)
+
+    elif method == "weighted":
+        return weighted_average(results, weight)
+
+    elif method == "detailed":
+        return pd.DataFrame(results)
+    else:
+        raise ValueError(f"Unknown method '{method}'. Available methods: 'majority', 'average', 'threshold', 'weighted', 'detailed'")
+
+