Blick in die Glaskugel

23.04.2018 | Blog Save time and costs with machine learning

Contracts or other legal documents are usually available in large quantities and in an unstructured form. Machine learning techniques help classify them, providing a clear ROI for the business.

ROI considerations are usually difficult, because afterwards it is difficult to prove with tangible numbers if the use of new software has really paid for itself. However, there are projects where the opposite is the case and which I therefore particularly like to implement for a client: text classification business cases.

Text classification as Supervised Machine Learning offers precise measurable time saving potential. The customer’s project manager quickly and reliably calculates the time savings and can clearly communicate the cost advantage to the buying center. This is a real business case, not just a use case.

Machine Learning and text classification reduce processing times

Text classification software is able to automatically assign documents to the right topic or category. For this purpose, metadata, for example a specific topic, is created for any text fragment and the document is then assigned to this topic on the basis of its content. The Machine Learning procedure takes over the sorting tasks. Whether it’s a real estate transaction, and thousands of documents need to be loaded into a virtual data room and sorted into a buyer-provided structure, or if hundreds of contracts need to be sorted through due diligence – in all of these situations, Machine Learning helps. It is a method of classifying and tagging numbers of data that is hard to handle manually or classifying it into a taxonomy.

Only a few training data required

The IntraFind Machine Learning procedure, which is used here, needs just a few training data and provides a very good assignment quality right from the start. We have eliminated the disadvantage of many learning methods that require large amounts of training documents. When training the learning process, the software also provides feedback, what should be changed in the training data, so that the system itself learns better. This also allows users who are not information scientists to use a technical learning process.
However, the system also ensures that its quality of assignment in productive use continues to improve: individual documents in which the AI ​​process is uncertain can be classified manually. This initiates a learning loop and the procedure is further trained and improved by this relevance feedback.

ROI of only a few months

The benefits of Machine Learning are particularly quick to detect in automatic text classification. A test on your data can be implemented without large effort and shows reliably what savings potential for the desired project the automatic sorting of data brings with it. The break even is a relatively small amount of documents. In our customer projects, we can prove ROI times of a few months. The Machine Learning process not only saves time and personnel costs, but also delivers better results due to the objective sorting work than the human and thus subjective and daily form-dependent sorting decisions

The author

Franz Kögl
Franz Kögl is co-founder and co-owner of IntraFind Software AG and has almost 20 years experience in Enterprise Search and Content Analytics.
Franz Kögl