How to tell a chatbot like ChatGPT or Sage to align texts in different languages in a table (and besides get a summary )

If you have the same text in several languages and you want the different versions neatly arranged in a table with the sentences in the different language versions next to each other, a chatbot like ChatGPT may be able to help. Especially when texts come in pdf format, just copying and pasting them can already be a nightmare what with the mess of useless linebreaks and word separations – let alone aligning the different language versions in a table.

I used the following prompt to extract and align English and German text from patent claims:

From the following English and German text, please copy all the sentences into a two-column table with the English and German version of the same sentence in one line, but different columns next to each other (please don’t omit the numbering).

It worked perfectly fine in ChatGPT-4 and Sage (both powered by OpenAI, they can be tested here), as long as the chunks of text I used weren’t too long. This is the alignment I obtained from ChatGPT-4:

English German
2. Control unit according to claim 1, characterised in that it has authentication means for verifying authorisation of the operator unit (70) for communication with the control unit (50; 150). 2. Steuergerät nach Anspruch 1, dadurch gekennzeichnet, dass es Authentifizierungsmittel zur Überprüfung einer Autorisierung des Bediengeräts (70) zu einer Kommunikation mit dem Steuergerät (50; 150) aufweist.
3. Control unit according to claim 1, characterised in that the authentication means are designed for verifying at least one access code (83) or identification code or password, which the control unit (50; 150) receives from the operator unit (70) via the operator unit interface (64). 3. Steuergerät nach Anspruch 2, dadurch gekennzeichnet, dass die Authentifizierungsmittel zur Überprüfung mindestens eines Zugangsschlüssels (83) oder Identifizierungscodes oder Passworts ausgestaltet sind, die das Steuergerät (50; 150) über die Bediengerät-Schnittstelle (64) von dem Bediengerät (70) empfängt.
4. Control unit according to any of the preceding claims, characterised in that the operator unit interface (64) or a part of the operator unit interface (64) is located onboard the control unit (50; 150), in particular inside a housing (65; 165) of the control unit (50; 150), and/or that it includes a vehicle electrical system interface (63) for communication with a vehicle electrical system (93) of the motor vehicle (90). 4. Steuergerät nach einem der vorhergehenden Ansprüche, dadurch gekennzeichnet, dass die Bediengerät-Schnittstelle (64) oder ein Teil der Bediengerät-Schnittstelle (64) an Bord des Steuergeräts (50; 150), insbesondere innerhalb eines Gehäuses (65; 165) des Steuergeräts (50; 150), angeordnet ist und/oder dass es eine Bordnetz-Schnittstelle (63) zur Kommunikation mit einem Bordnetz (93) des Kraftfahrzeugs (90) umfasst.

And then once you are at it, you can also ask the bot to give you a summary of the text, which by the way is more in line with what chatbots are all about. This is the summary Sage generated from a set of 14 patent claims:

This text describes a patent application for a control unit for a trailer coupling that can be mounted on a motor vehicle or on a trailer. The control unit has control and evaluation means for controlling and evaluating at least one function of the trailer coupling and at least one sensor signal. It also has an operator unit interface for communication with a mobile operator unit, such as a smartphone, and is designed to transmit status signals representing the status of the trailer coupling and to receive control signals for controlling the unit. The control unit is also designed to report the position of the coupling arm to the operator unit, which can then display the position on a graphical user interface. The text also describes several features of the control unit, including authentication means, control outputs, sensor inputs, and the ability to activate electrical contacts and output data received from the trailer. The text also describes a program module for the operator unit that can generate a graphical display and user interface based on the status signals received from the control unit and can generate control signals to control the unit.

It is important always to bear in mind that these systems have not been programmed to perform predefined tasks the way conventional software does. The same prompt may therefore lead to different results when repeated (which is what happened in the case of term extraction using ChatGPT).

Any further ideas about how AI can support us in our knowledge work? Feel free to share!

About the author

Anja Rütten is a freelance conference interpreter for German (A), Spanish (B), English (C), and French (C) based in Düsseldorf, Germany. She has specialised in knowledge management since the mid-1990s.

She is a full member of AIIC, an accredited freelance interpreter with the EU institutions and the European Patent Office, and has vast experience as a university lecturer.



, ,



Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.