{"observation":{"id":"a0e9e32e-85ea-493d-959a-ab26271ebc7f","tool":"retab","tool_name":"Retab","criterion":"table-and-record-completeness","criterion_name":"Table & Record Completeness","criterion_definition":"Are all tabular rows (transactions, line items) extracted without merging, duplication, omission, or phantom records?","criterion_evidence_type":"transformation","criterion_rank_role":"decisive","criterion_rank_role_reason":"For document data extraction, missing, merged, duplicated, or phantom rows directly corrupt the structured dataset and break downstream querying. (3 of 3 judges)","scenario":"bank-statement-pdf","scenario_name":"Bank Statement PDF","group_tag":"financial-document-extraction","scenario_description":"A 4-page bank statement PDF with 51 transactions, balances, rewards, and disclaimer text, used to test schema-driven extraction of dense financial tables and multi-page continuity.","modality":"pdf","input_text":null,"input_artifact_refs":[{"alt":null,"url":"https://d3epheqghktydj.cloudfront.net/extract-and-query-structured-data-from-d-bank-statement-pdf-25ec532be6e1.pdf","role":"input","filename":"Bank Statement PDF.pdf"}],"stresses":["Table extraction across 50+ transaction rows","Multi-page continuity with BALANCE FORWARD bridges","Structured metadata vs. free-text transaction descriptions","Numerical accuracy for balances, deposits, withdrawals, and summaries","Nested schema population for account, branch, balances, rewards, and disclaimers"],"verdict":"mixed","score":null,"score_total":null,"note":"Overcounts the transaction table summary: total_transactions is 43, while the report says the expected count is 40 after excluding Balance Forward, tax, and charge entries.","evidence_state":"verified","source":null,"artifacts":[{"url":"https://d3epheqghktydj.cloudfront.net/research-media-retab-bank-statement-total-transactions-c378712469f2.png","role":"output","alt":null},{"url":"https://d3epheqghktydj.cloudfront.net/research-media-docsumo-bank-statement-output-1-7228ed951c2f.xlsx","role":"context","alt":null}],"run_id":"ec4d736d-95f9-4c88-884c-e280435f7b7b","study_title":"Extract and query structured data from documents using natural language","study_kind":"generation","research_task":"86b9y25e5","tested_at":null,"completeness":"input-and-output","input":{"state":"files","text":null,"files":[{"url":"https://d3epheqghktydj.cloudfront.net/extract-and-query-structured-data-from-d-bank-statement-pdf-25ec532be6e1.pdf","filename":"Bank Statement PDF.pdf","alt":"Bank Statement PDF","role":"input"}],"modality":"pdf","stresses":["Table extraction across 50+ transaction rows","Multi-page continuity with BALANCE FORWARD bridges","Structured metadata vs. free-text transaction descriptions","Numerical accuracy for balances, deposits, withdrawals, and summaries","Nested schema population for account, branch, balances, rewards, and disclaimers"]},"tool_page_slug":"retab","tool_url":"https://aidemos.com/tools/retab","permalink":"https://aidemos.com/evidence/a0e9e32e-85ea-493d-959a-ab26271ebc7f","api_url":"https://ai.aidemos.com/v1/observations/a0e9e32e-85ea-493d-959a-ab26271ebc7f"},"peers":[{"id":"f09df643-ec8c-4318-9173-3099db75a12d","tool":"datalab","tool_name":"Datalab","verdict":"failed","score":null,"score_total":null,"note":"Merges adjacent statement rows into one overly long transaction description, breaking row boundaries on the 21 Jun example instead of keeping the two transactions separate.","artifact_count":2,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-bank-statement-21-jun-transactions-890dc245679d.png","evidence_url":"https://aidemos.com/evidence/f09df643-ec8c-4318-9173-3099db75a12d"},{"id":"3b66a1fa-5d29-4130-8e37-0c4cdc033160","tool":"extend-ai","tool_name":"Extend AI","verdict":"worked","score":null,"score_total":null,"note":"The extractor keeps transaction rows as separate records, and the report says it captured all 51 transactions without merging adjacent rows.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-extend-ai-extracted-transaction-item-4618ec1646be.png","evidence_url":"https://aidemos.com/evidence/3b66a1fa-5d29-4130-8e37-0c4cdc033160"},{"id":"76a18022-79e8-45a6-b3c9-bacd1c8b7330","tool":"landing-ai","tool_name":"Landing AI","verdict":"worked","score":null,"score_total":null,"note":"Preserves the full transaction table as 51 separate statement records, without merging rows or dropping entries.","artifact_count":5,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-landingai-bank-statement-output-d3ff94f60bcb.json","evidence_url":"https://aidemos.com/evidence/76a18022-79e8-45a6-b3c9-bacd1c8b7330"},{"id":"b9bfb6b9-ede4-4ea0-ab6e-68c2c5b3895b","tool":"llamaparse","tool_name":"LlamaParse","verdict":"failed","score":null,"score_total":null,"note":"Over-segments the bank-statement transaction table, outputting 54 transaction rows for a statement the report says contains 51 transactions and leaving the summary at 44 total_transactions, so row counts are not self-consistent.","artifact_count":3,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-llamaparse-bank-statement-transaction-co-ea2cee32bed3.png","evidence_url":"https://aidemos.com/evidence/b9bfb6b9-ede4-4ea0-ab6e-68c2c5b3895b"},{"id":"aabe6c7c-96c6-45f9-8357-03560e885eb4","tool":"nanonets","tool_name":"Nanonets","verdict":"failed","score":null,"score_total":null,"note":"It merges adjacent bank-statement rows into a single record description, so one extracted transaction can absorb neighboring content rather than staying row-bounded.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-bank-statement-18-jun-transactions-d294f17fc318.png","evidence_url":"https://aidemos.com/evidence/aabe6c7c-96c6-45f9-8357-03560e885eb4"},{"id":"d7084667-c138-440e-ae8a-85055f890e01","tool":"reducto","tool_name":"Reducto","verdict":"failed","score":null,"score_total":null,"note":"Does not preserve the repeating transaction table faithfully, returning 53 transaction records for a statement that contains 51 transactions.","artifact_count":3,"thumbnail":"https://d3epheqghktydj.cloudfront.net/research-media-reducto-transactions-5f1507b139e2.png","evidence_url":"https://aidemos.com/evidence/d7084667-c138-440e-ae8a-85055f890e01"},{"id":"1355f0ad-f96e-476c-9a0d-ab7e11352c1e","tool":"unstract","tool_name":"Unstract","verdict":"worked","score":null,"score_total":null,"note":"Keeps the transaction table intact across all four pages, with the full 51-row record set present and no merged or dropped rows.","artifact_count":2,"thumbnail":"https://d3epheqghktydj.cloudfront.net/unstract-unstract-bank-statement-output-f0bd422ffa9e.json","evidence_url":"https://aidemos.com/evidence/1355f0ad-f96e-476c-9a0d-ab7e11352c1e"}],"other_criteria":[{"id":"bb804c81-06cb-4d05-82a9-804c000f4321","criterion":"extraction-accuracy","criterion_name":"Extraction Accuracy","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Extracts key statement values as typed fields, including account holder MR SEENIVASAN, account number 42710540422, and opening and closing balances 114453.65 and 116149.46.","artifact_count":1,"evidence_url":"https://aidemos.com/evidence/bb804c81-06cb-4d05-82a9-804c000f4321"},{"id":"3cd1ce35-b4b0-4b71-9a49-f28d7f5e3622","criterion":"extraction-accuracy","criterion_name":"Extraction Accuracy","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Preserves long disclaimer prose in dedicated fields, capturing two separate strings for insurance_coverage and reporting_period rather than collapsing them into one blob.","artifact_count":1,"evidence_url":"https://aidemos.com/evidence/3cd1ce35-b4b0-4b71-9a49-f28d7f5e3622"},{"id":"18014a53-cc1f-4a5d-8552-fc026341e820","criterion":"schema-adherence","criterion_name":"Schema Adherence","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Reconstructs dense statement OCR into the requested nested JSON hierarchy, populating separate statement, account_holder, account, branch, balances, transactions, rewards, and disclaimers objects instead of flattening everything into text.","artifact_count":4,"evidence_url":"https://aidemos.com/evidence/18014a53-cc1f-4a5d-8552-fc026341e820"},{"id":"0067b41d-0138-4fc0-80b1-151bb93e2539","criterion":"semantic-field-enrichment","criterion_name":"Semantic Field Enrichment","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Derives transaction_type and transaction_id on transaction rows, classifying one record as UPI with transaction_id 917615251879 and cheque_number left empty.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/0067b41d-0138-4fc0-80b1-151bb93e2539"}],"appears_in":[{"page_type":"ranking","slug":"document-extraction","title":"Best AI Tools for Extracting Structured Data from PDFs and Business Documents","url":"https://aidemos.com/best/document-extraction","binding":"run"},{"page_type":"tool","slug":"unstract","title":null,"url":"https://aidemos.com/tools/unstract","binding":"run"}],"same_scenario":[{"id":"f09df643-ec8c-4318-9173-3099db75a12d","tool":"datalab","tool_name":"Datalab","verdict":"failed","score":null,"score_total":null,"note":"Merges adjacent statement rows into one overly long transaction description, breaking row boundaries on the 21 Jun example instead of keeping the two transactions separate."},{"id":"3b66a1fa-5d29-4130-8e37-0c4cdc033160","tool":"extend-ai","tool_name":"Extend AI","verdict":"worked","score":null,"score_total":null,"note":"The extractor keeps transaction rows as separate records, and the report says it captured all 51 transactions without merging adjacent rows."},{"id":"76a18022-79e8-45a6-b3c9-bacd1c8b7330","tool":"landing-ai","tool_name":"Landing AI","verdict":"worked","score":null,"score_total":null,"note":"Preserves the full transaction table as 51 separate statement records, without merging rows or dropping entries."},{"id":"b9bfb6b9-ede4-4ea0-ab6e-68c2c5b3895b","tool":"llamaparse","tool_name":"LlamaParse","verdict":"failed","score":null,"score_total":null,"note":"Over-segments the bank-statement transaction table, outputting 54 transaction rows for a statement the report says contains 51 transactions and leaving the summary at 44 total_transactions, so row counts are not self-consistent."},{"id":"aabe6c7c-96c6-45f9-8357-03560e885eb4","tool":"nanonets","tool_name":"Nanonets","verdict":"failed","score":null,"score_total":null,"note":"It merges adjacent bank-statement rows into a single record description, so one extracted transaction can absorb neighboring content rather than staying row-bounded."},{"id":"d7084667-c138-440e-ae8a-85055f890e01","tool":"reducto","tool_name":"Reducto","verdict":"failed","score":null,"score_total":null,"note":"Does not preserve the repeating transaction table faithfully, returning 53 transaction records for a statement that contains 51 transactions."},{"id":"1355f0ad-f96e-476c-9a0d-ab7e11352c1e","tool":"unstract","tool_name":"Unstract","verdict":"worked","score":null,"score_total":null,"note":"Keeps the transaction table intact across all four pages, with the full 51-row record set present and no merged or dropped rows."}]}