Interactive editing of a machine-generated document
Embodiments relate to interactive editing of a machine-generated document. A computer-implemented method includes receiving, by a processor, a machine-generated document and performing a comparison of a current state of the machine-generated document to a previous state. A user edit is identified as one or more user-replaced tokens of a previous token sequence based at least in part on the comparison. A new version of the machine-generated document is generated that includes the one or more user-replaced tokens and identifies one or more related tokens to replace with a suggested replacement token sequence associated with the one or more user-replaced tokens. A suggestion list is generated for display to the user in a graphical user interface to indicate the suggested replacement token sequence to replace the one or more related tokens.
1. A computer-implemented method comprising:
receiving, by an interactive editing tool executed by a processor, a machine-generated document that is generated by a generative model, the machine-generated document including a translation of a corpus in a programming language from a first representation to a second representation;
performing, by the interactive editing tool executed by the processor, a comparison of a current state of the machine-generated document to a previous state of the machine-generated document, wherein the comparison of the current state of the machine-generated document to the previous state is during an editing session of the machine-generated document through the interactive editing tool;
identifying, by the interactive editing tool executed by the processor, a user edit made to the machine-generated document as one or more user-replaced tokens of a previous token sequence in the machine-generated document based at least in part on the comparison;
identifying, by the interactive editing tool executed by the processor, one or more related tokens appearing after the one or more user-replaced tokens in the previous token sequence, wherein the one or more related tokens are identified based on the translation including the user edit incorporated using the generative model configured with language structure rules for the translation;
generating, by the interactive editing tool executed by the processor, or triggering generation, by an application, of a new version of the machine-generated document that includes the one or more user-replaced tokens and identifies the one or more related tokens to replace with a suggested replacement token sequence associated with the one or more user-replaced tokens;
generating, by the interactive editing tool executed by the processor, a suggestion list for display to the user in a graphical user interface to indicate the suggested replacement token sequence to replace the one or more related tokens;
outputting, by the interactive editing tool executed by the processor, the suggestion list in the graphical user interface during the editing session through the interactive editing tool;
confirming, by the interactive editing tool executed by the processor, user acceptance of the suggested replacement token sequence at one or more replacement locations in the new version of the machine-generated document during the editing session;
selecting, by the interactive editing tool executed by the processor, an alternate version of the machine-generated document as the current state based at least in part on the user rejecting one or more changes from the suggestion list during the editing session, to make the alternate version of the machine-generated document a basis for subsequent comparisons; and
continuing to monitor, by the interactive editing tool executed by the processor, for further edits to the machine-generated document through the graphical user interface after confirming the user acceptance and generating further suggested updates of the machine-generated document until the editing session is complete, wherein the further suggested updates comprise an alternate suggested replacement sequence based on a user rejection of one or more of the further suggested updates during the editing session.
2. The computer-implemented method of claim 1 , wherein the new version of the machine-generated document is generated using autoregression to produce a next token based at least in part on one or more previous tokens generated up to a current location within the machine-generated document.
3. The computer-implemented method of claim 1 , further comprising performing a multiple way difference comparison to track multiple changes to the machine-generated documents to identify edits.
4. The computer-implemented method of claim 1 , wherein the new version of the machine-generated document is generated using non-causal generation to search forward and backward from the one or more user-replaced tokens to replace the one or more related tokens with the suggested replacement token sequence associated with the one or more user-replaced tokens.
5. A computer system comprising:
a memory having computer readable instructions; and
one or more processors for executing the computer readable instructions of an interactive editing tool, the computer readable instructions configured to perform operations by the interactive editing tool comprising:
receiving a machine-generated document that is generated by a generative model, the machine-generated document including a translation of a corpus in a programming language from a first representation to a second representation;
performing a comparison of a current state of the machine-generated document to a previous state of the machine-generated document, wherein the comparison of the current state of the machine-generated document to the previous state is during an editing session of the machine-generated document through the interactive editing tool;
identifying a user edit made to the machine-generated document as one or more user-replaced tokens of a previous token sequence in the machine-generated document based at least in part on the comparison;
identifying one or more related tokens appearing after the one or more user-replaced tokens in the previous token sequence, wherein the one or more related tokens are identified based on the translation including the user edit incorporated using the generative model configured with language structure rules for the translation;
generating or triggering generation, by an application, of a new version of the machine-generated document that includes the one or more user-replaced tokens and identifies the one or more related tokens to replace with a suggested replacement token sequence associated with the one or more user-replaced tokens;
generating a suggestion list for display to the user in a graphical user interface to indicate the suggested replacement token sequence to replace the one or more related tokens;
outputting the suggestion list in the graphical user interface during the editing session through the interactive editing tool;
confirming user acceptance of the suggested replacement token sequence at one or more replacement locations in the new version of the machine-generated document during the editing session;
selecting an alternate version of the machine-generated document as the current state based at least in part on the user rejecting one or more changes from the suggestion list during the editing session, to make the alternate version of the machine-generated document a basis for subsequent comparisons; and
continuing to monitor for further edits to the machine-generated document through the graphical user interface after confirming the user acceptance and generating further suggested updates of the machine-generated document until the editing session is complete, wherein the further suggested updates comprise an alternate suggested replacement sequence based on a user rejection of one or more of the further suggested updates during the editing session.
6. The system of claim 5 , wherein the new version of the machine-generated document is generated using autoregression to produce a next token based at least in part on one or more previous tokens generated up to a current location within the machine-generated document.
7. The system of claim 5 , wherein the operations further comprise:
performing a multiple way difference comparison to track multiple changes to the machine-generated document to identify edits.
8. The system of claim 5 , wherein the new version of the machine-generated document is generated using non-causal generation to search forward and backward from the one or more user-replaced tokens to replace the one or more related tokens with the suggested replacement token sequence associated with the one or more user-replaced tokens.
9. A computer program product comprising a computer readable storage medium having program instructions embodied therewith of an interactive editing tool, the program instructions executable by one or more processors and configured to perform operations by the interactive editing tool comprising:
receiving a machine-generated document that is generated by a generative model, the machine-generated document including a translation of a corpus in a programming language from a first representation to a second representation;
performing a comparison of a current state of the machine-generated document to a previous state of the machine-generated document, wherein the comparison of the current state of the machine-generated document to the previous state is during an editing session of the machine-generated document through the interactive editing tool;
identifying a user edit made to the machine-generated document as one or more user-replaced tokens of a previous token sequence in the machine-generated document based at least in part on the comparison;
identifying one or more related tokens appearing after the one or more user-replaced tokens in the previous token sequence, wherein the one or more related tokens are identified based on the translation including the user edit incorporated using the generative model configured with language structure rules for the translation;
generating or triggering generation, by an application, of a new version of the machine-generated document that includes the one or more user-replaced tokens and identifies the one or more related tokens to replace with a suggested replacement token sequence associated with the one or more user-replaced tokens;
generating a suggestion list for display to the user in a graphical user interface to indicate the suggested replacement token sequence to replace the one or more related tokens;
outputting the suggestion list in the graphical user interface during the editing session through the interactive editing tool;
confirming user acceptance of the suggested replacement token sequence at one or more replacement locations in the new version of the machine-generated document during the editing session;
selecting an alternate version of the machine-generated document as the current state based at least in part on the user rejecting one or more changes from the suggestion list during the editing session, to make the alternate version of the machine-generated document a basis for subsequent comparisons; and
continuing to monitor for further edits to the machine-generated document through the graphical user interface after confirming the user acceptance and generating further suggested updates of the machine-generated document until the editing session is complete, wherein the further suggested updates comprise an alternate suggested replacement sequence based on a user rejection of one or more of the further suggested updates during the editing session.
10. The computer program product of claim 9 , wherein the operations further comprise:
performing a multiple way difference comparison to track multiple versions of the machine-generated document with divergent region alternatives.
11. The computer program product of claim 9 , wherein the new version of the machine-generated document is generated using autoregression to produce a next token based at least in part on one or more previous tokens generated up to a current location within the machine-generated document.
12. The computer program product of claim 9 , wherein the new version of the machine-generated document is generated using non-causal generation to search forward and backward from the one or more user-replaced tokens to replace the one or more related tokens with the suggested replacement token sequence associated with the one or more user-replaced tokens.