Topic modeling for conversations for mental health helplines with utterance embedding
Conversations with topics that are locally contextual often produces incoherent topic modeling results using standard methods. Splitting a conversation into its individual utterances makes it possible to avoid this problem. However, with increased data sparsity, different methods need to be consider...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-03-01
|
| Series: | Telematics and Informatics Reports |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772503024000124 |
| _version_ | 1797260259345039360 |
|---|---|
| author | Salim Salmi Rob van der Mei Saskia Mérelle Sandjai Bhulai |
| author_facet | Salim Salmi Rob van der Mei Saskia Mérelle Sandjai Bhulai |
| author_sort | Salim Salmi |
| collection | DOAJ |
| description | Conversations with topics that are locally contextual often produces incoherent topic modeling results using standard methods. Splitting a conversation into its individual utterances makes it possible to avoid this problem. However, with increased data sparsity, different methods need to be considered. Baseline bag-of-word topic modeling methods for regular and short-text, as well as topic modeling methods using transformer-based sentence embeddings were implemented. These models were evaluated on topic coherence and word embedding similarity. Each method was trained using single utterances, segments of the conversation, and on the full conversation. The results showed that utterance-level and segment-level data combined with sentence embedding methods performs better compared to other non-sentence embedding methods or conversation-level data. Among the sentence embedding methods, clustering using HDBScan showed the best performance. We suspect that ignoring noisy utterances is the reason for better topic coherence and a relatively large improvement in topic word similarity. |
| first_indexed | 2024-03-07T14:27:48Z |
| format | Article |
| id | doaj.art-28a8b1272d1148d9931ae850cf9461d4 |
| institution | Directory Open Access Journal |
| issn | 2772-5030 |
| language | English |
| last_indexed | 2024-04-24T23:22:29Z |
| publishDate | 2024-03-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Telematics and Informatics Reports |
| spelling | doaj.art-28a8b1272d1148d9931ae850cf9461d42024-03-16T05:10:01ZengElsevierTelematics and Informatics Reports2772-50302024-03-0113100126Topic modeling for conversations for mental health helplines with utterance embeddingSalim Salmi0Rob van der Mei1Saskia Mérelle2Sandjai Bhulai3Centrum Wiskunde & Informatica, Netherlands; Correspondence to: P.O. Box 94079, 1090 GB Amsterdam, Netherlands.Centrum Wiskunde & Informatica, Netherlands113 Suicide Prevention, NetherlandsVrije Universiteit Amsterdam, NetherlandsConversations with topics that are locally contextual often produces incoherent topic modeling results using standard methods. Splitting a conversation into its individual utterances makes it possible to avoid this problem. However, with increased data sparsity, different methods need to be considered. Baseline bag-of-word topic modeling methods for regular and short-text, as well as topic modeling methods using transformer-based sentence embeddings were implemented. These models were evaluated on topic coherence and word embedding similarity. Each method was trained using single utterances, segments of the conversation, and on the full conversation. The results showed that utterance-level and segment-level data combined with sentence embedding methods performs better compared to other non-sentence embedding methods or conversation-level data. Among the sentence embedding methods, clustering using HDBScan showed the best performance. We suspect that ignoring noisy utterances is the reason for better topic coherence and a relatively large improvement in topic word similarity.http://www.sciencedirect.com/science/article/pii/S2772503024000124Topic modelingSentence embeddingConversationsMental healthBert |
| spellingShingle | Salim Salmi Rob van der Mei Saskia Mérelle Sandjai Bhulai Topic modeling for conversations for mental health helplines with utterance embedding Telematics and Informatics Reports Topic modeling Sentence embedding Conversations Mental health Bert |
| title | Topic modeling for conversations for mental health helplines with utterance embedding |
| title_full | Topic modeling for conversations for mental health helplines with utterance embedding |
| title_fullStr | Topic modeling for conversations for mental health helplines with utterance embedding |
| title_full_unstemmed | Topic modeling for conversations for mental health helplines with utterance embedding |
| title_short | Topic modeling for conversations for mental health helplines with utterance embedding |
| title_sort | topic modeling for conversations for mental health helplines with utterance embedding |
| topic | Topic modeling Sentence embedding Conversations Mental health Bert |
| url | http://www.sciencedirect.com/science/article/pii/S2772503024000124 |
| work_keys_str_mv | AT salimsalmi topicmodelingforconversationsformentalhealthhelplineswithutteranceembedding AT robvandermei topicmodelingforconversationsformentalhealthhelplineswithutteranceembedding AT saskiamerelle topicmodelingforconversationsformentalhealthhelplineswithutteranceembedding AT sandjaibhulai topicmodelingforconversationsformentalhealthhelplineswithutteranceembedding |