{"id":23108886,"url":"https://github.com/archishmansengupta/qlamda","last_synced_at":"2026-04-28T11:03:59.286Z","repository":{"id":240103942,"uuid":"766606470","full_name":"ArchishmanSengupta/qlamda","owner":"ArchishmanSengupta","description":"qlamda: txt2ques generation model","archived":false,"fork":false,"pushed_at":"2024-05-16T17:04:24.000Z","size":783,"stargazers_count":2,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-04-03T23:14:50.489Z","etag":null,"topics":["bert-model","deep-learning","flashtext","huggingface","levenshtein-distance","nlp","python","sense2vec","t5-transformer","wordnet"],"latest_commit_sha":null,"homepage":"","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ArchishmanSengupta.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-03-03T18:13:01.000Z","updated_at":"2024-11-23T15:39:01.000Z","dependencies_parsed_at":"2024-05-16T19:28:53.116Z","dependency_job_id":"1d4cf3fd-49b8-4503-bde5-989ce6542c31","html_url":"https://github.com/ArchishmanSengupta/qlamda","commit_stats":null,"previous_names":["archishmansengupta/qlamda"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/ArchishmanSengupta/qlamda","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ArchishmanSengupta%2Fqlamda","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ArchishmanSengupta%2Fqlamda/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ArchishmanSengupta%2Fqlamda/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ArchishmanSengupta%2Fqlamda/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ArchishmanSengupta","download_url":"https://codeload.github.com/ArchishmanSengupta/qlamda/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ArchishmanSengupta%2Fqlamda/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32377599,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-28T09:24:15.638Z","status":"ssl_error","status_checked_at":"2026-04-28T09:24:15.071Z","response_time":56,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["bert-model","deep-learning","flashtext","huggingface","levenshtein-distance","nlp","python","sense2vec","t5-transformer","wordnet"],"created_at":"2024-12-17T01:31:48.479Z","updated_at":"2026-04-28T11:03:59.247Z","avatar_url":"https://github.com/ArchishmanSengupta.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"QLaMDA: txt2Que generation\n\n![Screenshot 2024-03-24 at 12 25 21 AM](https://github.com/ArchishmanSengupta/QLaMDA/assets/71402528/9a6026ff-d1d0-4cea-8504-a20a7da02ad8)\n\nAutomated Question Generation (AQG) is essential in NLP, with applications in education, as- sessment, and conversational agents. qlamda presents a comprehensive approach to automated text summarization and question generation using NLP techniques. By leveraging the T5 transformer model, Sense2Vec, and Sentence Transformers, the system extracts key information and generates rel- evant questions. The integration of the MMR algorithm enhances keyword selection, while WordNet and Sense2Vec generate distractors, facilitating multiple-choice question creation. The results demon- strate the system’s ability to produce concise summaries and generate contextually relevant questions, showcasing its potential for educational and information retrieval applications.\n\n## Methodology\nThe methodology encompasses several steps, in- cluding text summarization, keyword extraction, question generation, and distractor generation. Each step plays a crucial role in ensuring the ef- fectiveness of the AQG system.\n\n### 1. Text Summarization\nFor text summarization, the T5 transformer model was utilized. T5, short for ”(Text-to-Text Transfer Transformer)” [1], is a versatile model that can per- form a wide range of NLP tasks by converting them into a text-to-text format. The model was pre- trained on a large corpus of text data and fine-tuned specifically for summarization tasks. The summa- rization process involves encoding the input text with a special prefix (”summarize: ”) to guide the model towards generating a summary. The model’s output is then post-processed to ensure proper sen- tence capitalization and formatting.\n\n### 2. Keyword Extraction\nKeyword extraction was enhanced using the Mul- tipartiteRank algorithm, which ranks keywords based on their relevance and importance within the text. Additionally, Sense2Vec was employed to un- derstand the semantic meaning of words, allowing for the identification of keywords that are semanti- cally similar to the original text. This step is vital for ensuring that the generated questions are con- textually relevant.\n\n\u003c!-- [![\\\\ S(c_i) = (1 - \\lambda) + \\lambda \\cdot \\sum_{c_j \\in \\text{pred}(c_i)} \\frac{w_{ij} \\cdot S(c_j)}{\\sum_{c_k \\in \\text{succ}(c_j)} w_{jk}}](https://latex.codecogs.com/svg.latex?%5C%5C%20S(c_i)%20%3D%20(1%20-%20%5Clambda)%20%2B%20%5Clambda%20%5Ccdot%20%5Csum_%7Bc_j%20%5Cin%20%5Ctext%7Bpred%7D(c_i)%7D%20%5Cfrac%7Bw_%7Bij%7D%20%5Ccdot%20S(c_j)%7D%7B%5Csum_%7Bc_k%20%5Cin%20%5Ctext%7Bsucc%7D(c_j)%7D%20w_%7Bjk%7D%7D)](#_) --\u003e\n\n\u003cimg width=\"401\" alt=\"Screenshot 2024-05-16 at 10 17 41 PM\" src=\"https://github.com/ArchishmanSengupta/qlamda/assets/71402528/792901b9-90c0-44c0-857c-28a4610d650c\"\u003e\n\n\n### 3. Question Generation\nQuestion generation was also performed using the T5 model, but with a different fine-tuning ap- proach. The model was trained on the SQuAD (Stanford Question Answering Dataset) [2] to understand the context-answer-question format. Given a context and an answer, the model generates a question that could lead to the provided answer within the given context. This process is crucial for creating educational materials and quizzes.\n\n### 4. Distractor Generation\nDistractors for multiple-choice questions were gen- erated using two approaches: WordNet and Sense2Vec. WordNet was used to find synonyms and related words that could serve as distrac- tors, while Sense2Vec was used to find semantically similar words. The Maximal Marginal Relevance (MMR) algorithm [3] was applied to select a di- verse set of distractors that are both relevant to the context and distinct from each other.\n\u003cimg width=\"416\" alt=\"Screenshot 2024-05-16 at 10 17 52 PM\" src=\"https://github.com/ArchishmanSengupta/qlamda/assets/71402528/2a8edef6-4c31-4370-a39b-fd7cad08b1bd\"\u003e\n\n### 5. JSON formatted response\n\n![qlamda_response](https://github.com/ArchishmanSengupta/qlamda/assets/71402528/0f9f93f5-0d9a-4470-8bc7-c416aee07f2e)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Farchishmansengupta%2Fqlamda","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Farchishmansengupta%2Fqlamda","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Farchishmansengupta%2Fqlamda/lists"}