A Position-Based Method for the Extraction of Financial Information in PDF Documents
Bibliographic record
Abstract
Financial documents are omnipresent and necessitate extensive human efforts in order to extract, validate and export their content. Considering the high importance of such data for effective business decisions, the need for accuracy goes beyond any attempt to accelerate the process or save resources. While many methods have been suggested in the literature, the problem to automatically extract reliable financial data remains difficult to solve in practice and even more challenging to implement in a real life context. This difficulty is driven by the specific nature of financial text where relevant information is principally contained in tables of varying formats. Table Extraction (TE) is considered as an essential but difficult step for restructuring data in a handleable format by identifying and decomposing table components. In this paper, we present a novel method for extracting financial information by the means of two simple heuristics. Our approach is based on the idea that the position of information, in unstructured but visually rich documents - as it is the case for the Portable Document Format (PDF) - is an indicator of semantic relatedness. This solution has been developed in partnership with the Caisse de Depot et Placement du Québec. We present here our method and its evaluation on a corpus of 600 financial documents, where an F-measure of 91% is reached.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.009 | 0.008 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.008 | 0.010 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".