Exception in thread "JavaFX Application Thread" java.lang.StringIndexOutOfBoundsException: String index out of range: -20

[pnrobinson]

at org.monarchinitiative.hpo_case_annotator.gui.controllers.DiseaseCaseDataController.hpoTextMiningButtonAction(DiseaseCaseDataController.java:337) This is happening with many different texts, e.g.,

Besides the clavicle and skull dysplasia, short stature,
scoliosis, enamel hypoplasia, delayed eruption of de-
ciduous teeth, low nasal bridge, delayed mineralization of pubic bone, broad femoral head with short fem-
oral neck, hypoplastic iliac wing, syringomyelia and
special faces were also observed in CCD children.
Furthermore, hypertelorism was observed in all CCD
children, except Family_A_II1. Supernumerary teeth,
retention cysts and long second metacarpal were
observed in all CCD children, except Family_A_II1
and Family_B_I

[ielis]

Hi @pnrobinson , when I use a query text

Besides the clavicle and skull dysplasia, short stature, scoliosis, enamel hypoplasia, delayed eruption of deciduous teeth, low nasal bridge, delayed mineralization of pubic bone, broad femoral head with short femoral neck, hypoplastic iliac wing, syringomyelia and special faces were also observed in CCD children.
Furthermore, hypertelorism was observed in all CCD children, except Family_A_II1. Supernumerary teeth, retention cysts and long second metacarpal were observed in all CCD children, except Family_A_II1 and Family_B_I.

these terms are identified:

begin	end	Label
42	56	Short stature
57	67	Scoliosis
68	86	Hypoplasia of dental enamel
87	123	Delayed eruption of primary teeth
87	103	Delayed eruption of teeth
124	141	Depressed nasal bridge
...	...	...

The exception is thrown when code tries to highlight the Delayed eruption of teeth, which is overlapping with the previous term.

I think we have 2 issues here:

• How to highlight terms when >1 terms are mined from the same text
• It seems that we are receiving a parent (Delayed eruption of teeth) and child (Delayed eruption of primary teeth) terms here. Is that a desired behaviour? Is there a setting for SciGraph to only return the most specific term?

[pnrobinson]

I think we can ask SciGraph to return only the most specific term and that would be the desired behavior. The overlap issue is more difficult, but it might be easiest just to merge the overlapping terms and show both HPOs? Or to break it up and highlight the first and last word separately?

[ielis]

I agree that if two terms that are not ancestor/descendent are mined from the same text chunk, then the most specific term should be selected. I hope that's what we can ask SciGraph to do.

I also think that we should show merge the overlapping terms and show both HPOs. All is clear then, I'll open and address an appropriate issue in HpoTextMining.