A hybrid approach of Weighted Fine-Tuned BERT extraction with deep Siamese Bi – LSTM model for semantic text similarity identification

Viji, D.; Revathy, S.

doi:10.1007/s11042-021-11771-6

Cited by 39 publications

(17 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In the 1970s and 1980s, the question answering system based on knowledge base came into being, which transformed questions into structured query statements, inquired knowledge base and returned answers. Computation of semantic similarity between problems [12] belongs to short-text similarity calculation, which has always been a research hotspot in the field of NLP. Most of the key information exists in unstructured information.…”

Section: Research Statusmentioning

confidence: 99%

Intelligent Question Answering System Based on Domain Knowledge Graph

Hao¹,

Wu²,

Chen³

et al. 2023

Atlantis Highlights in Computer Sciences

View full text Add to dashboard Cite

This paper introduces an intelligent question answering system based on the domain knowledge graph of military battle cases. Through the collection and accumulation of military big data, we first build a domain knowledge graph for military battle cases, and then use natural language processing related technologies to understand natural language problems, mainly intention recognition and slot filling. On problem intent identification. In this paper, BERT+TextCNN model is proposed to realize the intention classification of questions. LAC is used to segment the natural language questions and extract the entities in the question sentence in slot filling. The answer is then retrieved from the knowledge graph. The test results show that the accuracy of question comprehension in the question set is more than 90%, and it can answer most of the questions in the field quickly and accurately.

show abstract

Section: Research Statusmentioning

confidence: 99%

Intelligent Question Answering System Based on Domain Knowledge Graph

Hao¹,

Wu²,

Chen³

et al. 2023

Atlantis Highlights in Computer Sciences

View full text Add to dashboard Cite

show abstract

“…We used an unbalanced dataset for our work. [31] In this paper, a hybrid framework that combines a Deep Siamese network Bi-LSTM model with weighted fine-tuned BERT extraction is developed.…”

Section: The Word Embeddings Produced By Elmo Are Created By the Long...mentioning

confidence: 99%

Survey on Aspect Based Sentimental Analysis using Deep Learning Techniques

Nandini¹,

Priya²,

Sruthi³

et al. 2023

Int. J. Res. Publ. Rev.

View full text Add to dashboard Cite

People People are increasingly using social media to communicate their views and ideas in daily life. Analysis, processing, interpretation, and inference of subjective texts with the sentiment are all steps in the sentiment analysis process. Businesses utilise sentiment analysis for a variety of purposes, including market research, brand reputation analysis, customer experience evaluation, and social media influence research. It can be classified into document, sentence, and aspect-based types in accordance with the various needs for aspect granularity. In order to determine the aspects of entities and the sentiment conveyed for each one, Aspect Based Sentimental Analysis (ABSA), a fine-grained task in sentiment analysis, is performed. Since it offers more thorough and detailed data, ABSA has drawn a lot of interest in recent years. Product dependence, target dependence, and aspect dependence are the three fundamental difficulties ABSA faces. In order to choose the best aspect extraction approach for the proposed model, we first compared the accuracy of two different methods, TF-IDF and PMI. The precise extraction technique is examined first with Bert and subsequently with various classifiers. The many classifiers are finally combined to produce a superior accuracy. Pretrained BERT that has been fine-tuned performs very well on ABSA. It makes use of BERT intermediate layers to improve the effectiveness of BERT finetuning. The suggested approach will be tested using two datasets, Amazon Unlocked Mobile and tripadvisor hotel reviews, where hotel reviews produce more accurate findings.

show abstract

“…For the text prediction problem, its essence is next generation, that is, according to the previous content of the text, feature extraction and learning are performed, and by sharing weights, each layer can perform the same extraction task, which greatly improves the overall efficiency of the model. The RNN network can capture the information features of the previous text to a certain extent and provide predictions for the subsequent text according to the semantic relevance [5] . In this direction, the typical application is based on part-of-speech prediction, that is, by evaluating the part-of-speech of the input text one by one, predicting the part-of-speech of the subsequent texts to appear, and extracting words with high probability from many phrase libraries when making predictions.…”

Section: The Application Status Of Rnn In Text Predictionmentioning

confidence: 99%

Application research of RNN in text intelligent prediction scenario

Duan,

cheng,

Qiu

et al. 2023

Second International Conference on Electronic Information Technology (EIT 2023)

View full text Add to dashboard Cite

Since deep learning has been continuously developing, research on language models utilizing deep learning has increased, and certain results have been obtained. There are many research results, but RNN has unique advantages regarding the construction of language models. In addition, due to its weight-sharing cyclic network form, the language model can have the information memory function described above, which facilitates the prediction of content in the future. Text generation is one of the most common applications in this area. However, RNN may have some limitations, including its inability to remember the above for a prolonged period of time, making the language model formed by RNN relatively weak in terms of intelligence and ability to connect the languages under study. Therefore, this paper proposes to build a language prediction model based on the RNN network model LSTM. This will allow the language model to capture the previous characteristics better, resulting in improved intelligent text prediction.

show abstract

A hybrid approach of Weighted Fine-Tuned BERT extraction with deep Siamese Bi – LSTM model for semantic text similarity identification

Cited by 39 publications

References 37 publications

Intelligent Question Answering System Based on Domain Knowledge Graph

Intelligent Question Answering System Based on Domain Knowledge Graph

Survey on Aspect Based Sentimental Analysis using Deep Learning Techniques

Application research of RNN in text intelligent prediction scenario

Contact Info

Product

Resources

About