This document is part of Annotating Metabolic Processes.
This task requires some proficiency in using Argo. Please make sure that you know how to:
- set up component parameters,
- run workflows,
- track the progress of processing workflows and open the Manual Annotation Editor,
- use the Manual Annotation Editor.
Most of these actions are demonstrated in the step-by-step instructions below; however, you are encouraged to visit the tutorials page before continuing.
For the purpose of evaluating the usability of Argo, there are three workflows prepared for the curators:
- Manual annotation: This workflow reads PubMed abstracts and opens a Manual Annotation Editor for a curator to tag the entities of interest without any support from automatic processing available in Argo.
- Input: PubMed abstract identifiers.
- Output: Annotation files in XMI (interchangeable) format.
- Automatic annotation: This workflow reads PubMed abstracts and performs automatic recognition of the entities of interest. This workflow is purely automatic and does not involve any manual intervention from the curators. The objective of this workflow is to automatically “pre-annotate” the input abstracts for later manual inspection.
- Input: PubMed abstract identifiers.
- Output: Annotation files in XMI format.
- Manual correction: This workflow reads files that already contain annotations (coming from the second, automatic workflow) and opens a Manual Annotation Editor for a curator to correct (remove, add, modify) the automatically recognised annotations.
- Input: Annotation files in XMI format.
- Output: Annotation files in XMI format.
Although the second and third workflow could be combined into a single workflow, for the purpose of evaluation they are defined separately.
The three workflows are publicly available in Argo.
For convenience, you are advised to open Argo’s main application in a separate window and keep these instructions and the annotation guidelines handy.
A. Initial setup
The initial setup involves signing in to or registering an account with Argo, as well as creating folders that will be used to output annotations to. You will also make copies of publicly available, read-only workflows so that you can make some adjustments on them.initial setup step-by-step instructions
Step 1. Sign in
If you have not done so yet, create an account in Argo and sign in to it. Although it is possible to use Argo without registering, any user-created data (workflows and documents) will be automatically deleted at the end of a visit.
Tip: Registering an account in Argo requires a valid email address for verification that is sent immediately after creating an account. Please check your email spam/junk folders if you do not receive the verification email within minutes.
Step 2. Set up output folders
Each of the three workflows results in creating files containing annotations. Create the following three folders in your account’s Documents view:
- Manual annotation
- Automatic annotation
- Manual correction
Step 3. Create copies of the public workflows
The three workflows are publicly available for reading only. In order to be able to change their settings (for example, selecting an output folder for generated files) you have to make copies of these workflows.
Edit each of the workflows and save its copy. Although it is possible to have multiple workflows with the same name in Argo, we advise you to change the names of copies by, for example, preceding each of them with the “Copy of” prefix.
B. Manual annotation
This phase involves configuring the manual annotation workflow by setting up input PubMed IDs and folders for output annotations as well as running the workflows and performing the manual annotation.manual annotation step-by-step instructions
Step 4. Manual annotation: Change the workflow’s settings
The first workflow that you will work with is the manual annotation workflow. This workflow takes a number of PubMed abstract identifiers as input and generates annotation XMI files on its output.
- Set up the list of PubMed IDs of the PubMed Abstract Reader component (the first one in this workflow) to the first 5 IDs shown below.
- Set up the output folder of the XMI Writer component (the last one in this workflow) to the “Manual annotation” folder you created in Step 2.
PubMed abstract identifiers for the manual annotation workflow:
- 10502406 11284709 11393269 11590700 12547826
- 12821135 12972033 14709628 15126328 15547048
- 16081671 16216874 16293890 16479044 16601807
- 16980304 17520698 17698513 17977830 18664505
- 18759268 19161989 19442645 19786063 20155626
- 20467561 20615392 20876576 21468576 21804305
Note: Although it is possible to supply the entire PubMed ID list into the PubMed Abstract Reader component, it is not advisable. Argo will not save any of the annotated documents to the output folder before they are all complete. This comes from the fact that Argo allows you to revisit annotated documents and keeps the changes in an internal temporary location until you decide that all documents are complete (that is, there is no more changes to be done). Since the document set if fairly large, it may take days to annotate all of them which carries the risk of losing your work if Argo encounters technical problems. It is therefore advisable to annotate this set in smaller chunks.
Step 5. Manual annotation: Run the workflow and perform manual annotation
Run the manual annotation workflow and launch the Manual Annotation Editor. The editor will allow you to mark mentions of the entities of interest. The entity types present in the editor to choose from are Chemical, GeneOrGeneProduct, and Process. All of them have exactly the same features (or attributes). The begin and end features will be set automatically once you have marked a span of text. The id feature, however, has to be set manually. This feature is an external resource identifier of the marked span of text as defined in the guidelines.
Once the manual annotation is done, all output XMI files should be visible in the “Manual annotation” folder.
Tip: The Manual Annotation Editor window can be closed and reopened multiple times without losing any previous annotations. This is useful if you decide to take a longer break and close your browser. Next time, simply go back to the running workflow (in the Processes view) and click on the Launch Manual Annotation Editor button (as shown in the slides above).
C. Automatic Annotation
This phase involves configuring the input and output annotation folders of the automatic annotation workflow and running it.automatic annotation step-by-step instructions
Step 6. Automatic annotation: Change the workflow’s settings
The second workflow you will run is the automatic annotation workflow. Similarly to the previous workflow, this workflow takes a number of PubMed abstract identifiers as input and generates annotation XMI files on its output.
- Set up the list of PubMed IDs of the PubMed Abstract Reader component (the first one in this workflow) to the first 5 IDs shown below (note that they are different from those in Step 4).
- Set up the output folder of the XMI Writer component (the last one in this workflow) to the “Automatic annotation” folder you created in Step 2.
PubMed abstract identifiers for the automatic annotation workflow:
- 11097864 11375903 11453994 11723127 12635099
- 12960407 1435739 15115889 15135306 15955068
- 16156861 16236141 16426572 16563290 1666626
- 17483544 17619801 17848950 18515279 18729198
- 19053182 19221000 19464575 19805354 20187624
- 20493829 20705923 20937724 21743969 22046279
Step 7. Automatic annotation: Run the workflow
Run the automatic annotation workflow. The XMI files containing the automatically generated annotations will be placed in the “Automatic annotation” folder.
D. Manual correction
This phase is similar to the manual annotation phase. The difference lies in the source of input annotations which comes from the output of the previous, automatic annotation workflow.manual correction step-by-step instructions
Step 8. Manual correction: Change the workflow’s settings
The third workflow you will work with is the manual correction workflow. This workflow takes a number of annotation files in XMI format as input (in this case those produced by automatic annotation in Step 7) and generates annotation XMI files on its output.
- Set up the input folder of the XMI Reader component (the first one in this workflow) to the “Automatic annotation” folder.
- Set up the output folder of the XMI Writer component (the last one in this workflow) to the “Manual correction” folder you created in Step 2.
Step 9. Manual correction: Run the workflow and perform manual correction
Run the manual correction workflow and launch the Manual Annotation Editor. The editor will allow you to correct the annotations generated by the automatic annotation workflow. Following the annotation guidelines, you can add missed annotations, remove spuriously recognised ones, change labels (e.g., from Chemical to GeneOrGeneProduct), change spans and change values of the id feature as necessary.
Once the manual correction is done, all output XMI files should be visible in the “Manual correction” folder.
Frequently Asked Questions
- Can I try the workflows with other PubMed IDs for practice?
- Yes, you may enter any (preferably pertaining to metabolic processes) PubMed IDs as input to the workflows. Alternatively you can select some from the list below:
- 21223739 22064374 22585829 3037304 7906730
- 8076368 8445979 8558417 8915012 9972481
- What should I do if a recognised entity cannot be identified in ChEBI/UniProt/CTD?
- Nothing. It is expected that some entities will be new to the external dictionaries. They will be automatically pulled from the annotations in post-processing (not part of this curation task) and will constitute a set of new entries for the external dictionaries.
- How can I view or change annotations at a later visit?
- As long as the workflow with a Manual Annotation Editor is running, you can view or make changes to any of the documents that are part of this workflow by switching to the Processes view, identifying the right process, and clicking on the Launch Manual Annotation Editor button.
- If the process has finished, you can still view or change annotations; however, you have to create or edit another workflow. Notice that the “Metabolic processes: Manual correction” workflow is exactly what is needed to accomplish it. Simply edit this workflow, create a copy, rename it to something memorable, and change the input of the XMI Reader to the files/documents you want to view/edit. For the XMI Writer, if you wish to overwrite the input files, select the same folder and tick the “overwrite” checkbox.
- If you only want to view the files, remove the XMI Writer component from the workflow altogether.
- Can I see if XMI files contain annotations in the Documents view?
- XMI files contain annotations even if you have not created any explicitly; they include document metadata. Currently, the Documents view does not visually differentiate between XMI files with “default” annotations and those with added annotations.
- Can I download the annotations?
- Yes. Each of the workflows includes the XMI Writer component that writes the original text and annotations to your document space. Simply switch to the Documents view, locate and select the documents you wish to download and click on the Download button in the toolbar.
- Note: Currently you may download only one document at a time.