Summary

This is the source code for the paper "A Case Study and Qualitative Analysis of Simple Cross-Lingual Opinion Mining", which was accepted as full paper with oral presentation at the 13th International Conference on Knowledge Discovery and Information Retrieval (KDIR).

Workflow

Fig. 1. Plain text is first tokenized into sentences and passed to topic modeling and sentiment analysis. Topic modeling involves (1) converting sentences of both languages into embeddings with XLING, (2) clustering all embeddings with K-means and (3) deriving a topic label of each cluster. Sentiment analysis is performed using Textblob. Topic and sentiment scores are aggregated for the analysis.

Documentation

• XLING-simple-example.ipynb: shows simple examples for converting English and German sentences into sentence embeddings and the cosine similarity between them.
• article_based_topic_modeling_review.ipynb: demonstrates how sentences in a single article are clustered into different topics and summarised the corresponding sentiment distribution.
• cosine_similarity.ipynb: analyses the distribution of cosine similarity of sentences per topic.
• create_article_wise_csv.ipynb: creates data file which contains sentiments and topic assignment for every sentence in a single article for further analysis.
• radar_plot_final.ipynb: displays topic distribution per data source and document type.
• radarfactory.py: contains Python class for generating radar plot.
• regenerate_sentences_metadata_json.ipynb: helps to gather all related measurements and create an overall data file for analysis.
• sankey_plot_final.ipynb: generates the final sankey plot which shows the flow of topic distribution for increasing number of topics for a fixed among of input sentences.
• sankey_plot_k_clusters.ipynb: creates the first version of sankey plots.
• sentence_posting_time.ipynb: handles minor issues on sentences posting time.
• senti_util.py: includes utility class for sentiment analysis.
• sentiment.py: is a Python class for assigning SentiWordNet sentiment.
• sentiwordnet_vs_textblob.ipynb: shows distributions of Textblob sentiment and SentiWordNet sentiment.
• simple_distribution_of_sentiment.ipynb: includes detail version of sentiment distribution.
• time_related_distribution_of_sentiment.ipynb: indicates the change of sentiment of news and that of the responses from readers.
• top_sentences_per_clusters.ipynb: lists the top sentences per topic.
• top_words.ipynb: records top words per topic.
• util.py: is a Python utility class with simple functions for analysis.