[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-195219-105":53,"doc-detail-195219-en":127},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":120,"head_meta":122,"extra_data":124,"updated_unix":126},105,"en","restaurant-review-sentiment-analysis-and-relation-extraction-patched-models-evaluation","Restaurant Review Sentiment Analysis and Relation Extraction - Patched Models Evaluation","","Restaurant review processing combines sentiment analysis and relation extraction by using rule-based overrides and feature-based prompts. The material defines conditional labeling templates for aspects such as service, food, and ambience, including positive/negative overrides and feature-triggered interpretations. Multiple model variants are compared across override application, invariance behavior, and performance on Yelp-style targets, followed by detailed patch impact tables covering patch correctness, patch conditions, and patch consequents.",{"@graph":63,"@context":119},[64,80,102],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":11,"@type":70,"position":76},"https://docshare.wps.com/template/presentations/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/restaurant-review-sentiment-analysis-and-relation-extraction-patched-models-evaluation/195219/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/restaurant-review-sentiment-analysis-and-relation-extraction-patched-models-evaluation/195219.png","ImageObject",442,249,{"name":88,"@type":89},"Aurelia","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-10-07","2026-09-03",true,{"@type":98,"interactionType":99,"userInteractionCount":101},"InteractionCounter",{"@type":100},"ViewAction",8,{"@type":103,"mainEntity":104},"FAQPage",[105,111,115],{"name":106,"@type":107,"acceptedAnswer":108},"What is the role of override rules in the sentiment labeling process?","Question",{"text":109,"@type":110},"Override rules assign positive or negative labels based on explicit conditions in the review, such as an aspect being good or bad, or the presence of specific words.","Answer",{"name":112,"@type":107,"acceptedAnswer":113},"How does the evaluation compare original, regex, prompt, and patched model variants?",{"text":114,"@type":110},"The document reports metrics for correct patch application, invariance (non-application), and performance measures like F1, including results across different Yelp-related targets.",{"name":116,"@type":107,"acceptedAnswer":117},"What do patch conditions and patch consequents tables measure?",{"text":118,"@type":110},"Patch conditions table evaluates how specific relational facts (e.g., one person being the son/daughter or widow of another) affect differential outcomes, while patch consequents table quantifies the performance impact of resulting labels like food/service being good or bad.","https://schema.org",{"og:url":78,"og:type":121,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":123,"canonical":78},"index,follow",{"doc_id":125,"site_id":56},195219,1788446454,{"code":4,"msg":5,"data":128},{"doc_id":125,"user_id":129,"nickname":88,"user_avatar":130,"doc_module":9,"category_id":8,"category_name":11,"doc_title":59,"doc_description":61,"doc_content":131,"file_id":132,"file_url":133,"file_type":134,"file_size":135,"view_count":101,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":25,"language":136,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":137,"faqs":138,"seo_title":139,"seo_description":61,"update_tm":126,"read_time":47},1099514068365,"https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068","| Template |  | Examples |  |\n| --- | --- | --- | --- |\n| Override: If aspect is good, then label is positive |  | e0 : If service is good, then label is positive\u003Cbr>e1 : If food is good, then label is positive |  |\n| Patches | Override: If aspect is bad, then label is negative | e2 : If service is bad, then label is negative\u003Cbr>e3 : If ambience is bad then label is negative |  |\n| Override: If review contains words like word, then label is positive |  | e4 : If review contains words like zubin, then label is positive\u003Cbr>e5 : If review contains words like excellent, then label is positive |  |\n| Override: If review contains words like word, then label is negative |  | e6 : If review contains words like wug, then label is negative\u003Cbr>e7 : If review contains words like really bad, then label is negative |  |\n|  |  | e8 : If food is described as above average, then food is good |  |\n| Feature Based: If aspect is described |  | e9 : If food is described as wug, then food is bad |  |\n| as word, then aspect is good / bad |  | e10 : If food is described as zubin, then service is goode11 : If service is described as not great, then service is bad |  |\n| The aspect at the restaurant was |  | The service at the restaurant was really good. e0 , e3 |  |\n| adj |  | The food at the restaurant was wug. e6 , e9 |  |\n| Inputs The restaurant had adj aspect |  | The restaurant had really bad service. e7 , e2 , e11\u003Cbr>The restaurant had zubin ambience. e4 , e10 |  |\n| The aspect1 was adj1, the aspect2 |  | The food was good, the ambience was bad. e1 , e3 , e1 |  |\n| was adj2 |  | The service was good, the food was not good. e0 , e1 |  |\n| The aspect1 was adj1 but the aspect2 was really adj2 |  | The food was good, but the service was really bad. e7 , e1 , e0 The ambience was bad, but the food was really not wug. e3 , e9 |  |\n| The aspect1 was really adj1 eventhough aspect2 was adj2 |  | The food was really bad even though the ambience was excellent. e5 , e7 , The food was really zubin, even though the service was bad e4 , e10 , e0 | e8 |\n\n| Model | Sentiment Analysis |  |  |  | Relation Extraction |  |\n| --- | --- | --- | --- | --- | --- | --- |\n|  | Override | O-Inv | Feat | Feat-Inv | Feat | Feat-Inv |\n| ORIG | 50.0 | n/a | 59.1 | n/a | 14.5 | n/a |\n| ORIG+PF | 50.0 | n/a | 59.9 | n/a | 35.8 | n/a |\n| REGEX | 50.0 | 100.0 | 59.9 | 100.0 | 45.8 | 88.1 |\n| PROMPT | 68.7 | 63.8 | 64.3 | 85.4 | 13.9 | 87.6 |\n| PATCHED | 100.0 | 100.0 | 100.0 | 100.0 | 47.2 | 92.6 |\n\n\n| Model | Correctly patched (applies) | Invariance (does not apply) |\n| --- | --- | --- |\n| ORIG | 91.5 | n/a |\n| ORIG+PF | 91.1 | n/a |\n| REGEX | 91.0 | 99.5 |\n| PROMPT | 92.3 | 98.4 |\n| PATCHED | 95.8 | 99.4 |\n\n\n| Model | Correctly patched (applies) | Invariance (does not apply) |\n| --- | --- | --- |\n| ORIG | 52.1 | n/a |\n| PROMPT | 55.7 | 97.6 |\n| ORIG+PF | 53.5 | n/a |\n| REGEX | 55.1 | 100.0 |\n| PATCHED | 79.6 | 99.4 |\n\n\n| Model | Yelp-Stars | Yelp-Colloquial | Yelp-Colloquial-Control | WCR |\n| --- | --- | --- | --- | --- |\n| ORIG | 93.1 | 89.1 | 100.0 | 89.6 |\n| ORIG+PF | 93.6 | 88.6 | 100.0 | 88.9 |\n| REGEX | 92.7 | 91.9 | 88.1 | 90.0 |\n| PROMPT | 90.8 | 85.2 | 70.1 | 88.3 |\n| PATCHED | 94.5 | 93.2 | 100.0 | 90.1 |\n\n\n| Model |  | F1 |\n| --- | --- | --- |\n| ORIG |  | 65.5 |\n| ORIG+PF |  | 61.4 |\n| REGEX |  | 61.0 |\n| PROMPT |  | 65.7 |\n| PATCHED |  | 72.9 |\n| Using single patch | If p2 is the son of p1 , then label is negative\u003Cbr>If p1 is the son of p2 , then label is negative\u003Cbr>If p1 and p2 have a daughter, then label is positive If p1 and p2 have a son, then label is positive If p1 is the widow of p2 , then label is positive If p1 is the daughter of p2 , then label is negative\u003Cbr>If p2 is the daughter of p1 , then label is negative | 57.5\u003Cbr>58.9\u003Cbr>61.9\u003Cbr>66.8\u003Cbr>63.6\u003Cbr>50.7\u003Cbr>49.4 |\n\n\n| Patch Condition | Diff | Diff \\ Correct |\n| --- | --- | --- |\n| p2 is the son of p1 | 0.0 | NaN (0/0) |\n| p1 is the son of p2 | 75.0 | 75.0 |\n| p1 and p2 have a daughter | 63.3 | 93.8 |\n| p","cbCaicYG9gADU2YJ","https://ap.wps.com/l/cbCaicYG9gADU2YJ","pdf",995858,"English","# Overview\n## Override and feature-based labeling templates\n## Model comparison metrics and patch performance\n## Patch correctness, invariance, and target datasets\n## Patch condition and patch consequent evaluation","[{\"question\":\"What is the role of override rules in the sentiment labeling process?\",\"answer\":\"Override rules assign positive or negative labels based on explicit conditions in the review, such as an aspect being good or bad, or the presence of specific words.\"},{\"question\":\"How does the evaluation compare original, regex, prompt, and patched model variants?\",\"answer\":\"The document reports metrics for correct patch application, invariance (non-application), and performance measures like F1, including results across different Yelp-related targets.\"},{\"question\":\"What do patch conditions and patch consequents tables measure?\",\"answer\":\"Patch conditions table evaluates how specific relational facts (e.g., one person being the son/daughter or widow of another) affect differential outcomes, while patch consequents table quantifies the performance impact of resulting labels like food/service being good or bad.\"}]","Restaurant Review Sentiment Analysis and Relation Extraction - Patched Models Evaluation | PDF"]