Intellix -- End-User Trained Information Extraction for Document Archiving

Schuster, Daniel; Muthmann, Klemens; Esser, Daniel E.; Schill, Alexander; Berger, Michael; Weidling, Christoph; Aliyev, Kamil; Hofmeier, Andreas

doi:10.1109/icdar.2013.28

Cited by 79 publications

(41 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the more specific task of extracting information from business documents several works use a pattern matching approach. Schuster et al [2013], Rusinol et al [2013] and Cesarini et al [2003] require users to annotate which words should be extracted for a given document template, then automatically generate patterns matching those words. At test time, these patterns generate candidate words, which are scored using heuristics.…”

Section: Related Workmentioning

confidence: 99%

Attend, Copy, Parse End-to-end Information Extraction from Documents

Palm¹,

Laws²,

Winther

2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Document information extraction tasks performed by humans create data consisting of a PDF or document image input, and extracted string outputs. This end-to-end data is naturally consumed and produced when performing the task because it is valuable in and of itself. It is naturally available, at no additional cost. Unfortunately, state-of-the-art word classification methods for information extraction cannot use this data, instead requiring word-level labels which are expensive to create and consequently not available for many real life tasks. In this paper we propose the Attend, Copy, Parse architecture, a deep neural network model that can be trained directly on end-toend data, bypassing the need for word-level labels. We evaluate the proposed architecture on a large diverse set of invoices, and outperform a state-of-the-art production system based on word classification. We believe our proposed architecture can be used on many real life information extraction tasks where word classification cannot be used due to a lack of the required word-level labels. 1

show abstract

Section: Related Workmentioning

confidence: 99%

Attend, Copy, Parse End-to-end Information Extraction from Documents

Palm¹,

Laws²,

Winther

2019

2019 International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

show abstract

“…Flexible template-based extraction systems [1], [6]- [8] locate the required text in the document by using the distance and direction from important surrounding keywords such as field labels. Cesarini et al [6] only look at the nearest keyword whereas d'Andecy et al [7] computes distances and angles from every other word in the document and predicts the final location by averaging over the distances and angles from all words weighted by their itf-df scores.…”

Section: Related Workmentioning

confidence: 99%

One-Shot Template Matching for Automatic Document Data Capture

Dhakal¹,

Munikar²,

Dahal³

2019

2019 Artificial Intelligence for Transforming Business and Society (AITB)

View full text Add to dashboard Cite

In this paper, we propose a novel one-shot templatematching algorithm to automatically capture data from business documents with an aim to minimize manual data entry. Given one annotated document, our algorithm can automatically extract similar data from other documents having the same format. Based on a set of engineered visual and textual features, our method is invariant to changes in position and value. Experiments on a dataset of 595 real invoices demonstrate 86.4% accuracy.

show abstract

“…We adopted forty-year-old data frames [21]. Our example-based approach for user interaction has some similarities with the end-user-provided training examples used commercially for scanned business documents [22]. Some aspects of our templates, like the use of literals and semantic tags, were anticipated in [23].…”

Section: Prior Workmentioning

confidence: 99%

Green Information Extraction from Family Books

Nagy

2019

SN COMPUT. SCI.

View full text Add to dashboard Cite

There is rising demand for the retrieval of genealogical information from semi-structured books. Nuggets of personal interest are currently transcribed piecemeal by volunteers. GreenBook is an effective alternative for recovering most of the desired information from any of the hundreds of thousands of books of ancestral records that have already been scanned and digitized. It minimizes human intervention by letting the user benefit from automatically compiled statistics from the book, and lets computer search benefit from user insights. GreenBook combines enhanced template matching with spreadsheet-based interaction for the rapid specification of the text to be extracted. The accuracy and completeness of the extracted information are limited only by the user's stamina. About 3 h of user interaction yielded 99% precision and 97% recall on books of Scottish birth and marriage records, Ohio funeral home records, and a family history spanning 300 years. The system is designed to facilitate transition to new books.

show abstract

Intellix -- End-User Trained Information Extraction for Document Archiving

Cited by 79 publications

References 16 publications

Attend, Copy, Parse End-to-end Information Extraction from Documents

Attend, Copy, Parse End-to-end Information Extraction from Documents

One-Shot Template Matching for Automatic Document Data Capture

Green Information Extraction from Family Books

Contact Info

Product

Resources

About