Module tf.ner.show
Rendering of corpus material with annotations.
Expand source code Browse git
"""Rendering of corpus material with annotations.
"""
from itertools import chain
from ..advanced.helpers import dh
from ..browser.html import H
from ..browser.ner.websettings import (
repIdent,
repSummary,
SORT_DEFAULT,
SORTKEY_DEFAULT,
SORTDIR_DEFAULT,
SORTDIR_DESC,
LIMIT_BROWSER,
LIMIT_NB,
)
from .helpers import toAscii
from .settings import SET_SHEET
class Show:
def showEntityOverview(self):
"""Generates HTML for an overview of the entities.
The entity overview consists of a listing of the possible
entity kinds with for each kind how many entities there are of that kind.
Returns
-------
string or void
If called by the browser, it returns the HTML string.
Otherwise, it displays the HTML string in the output, assuming
it is a cell in a Jupyter Notebook.
"""
settings = self.settings
keywordFeatures = settings.keywordFeatures
browse = self.browse
setData = self.getSetData()
content = H.p(
H.span(
H.code(f"{len(es):>5}"),
" x ",
H.span(repSummary(keywordFeatures, fVals)),
)
+ H.br()
for (fVals, es) in sorted(
setData.entitySummary.items(), key=lambda x: (-len(x[1]), x[0])
)
)
if browse:
return content
dh(content)
def showEntities(
self, activeEntity=None, sortKey=None, sortDir=None, cutOffFreq=None
):
"""Generates HTML for a sorted list of the entities.
The entity list consists of a table of entity identifiers, with the columns for
the kind and frequency of the entities.
There may be one active entity, and that one will be highlighted.
Parameters
----------
activeEntity: tuple, optional None
The entity that must be highlighted, given as entity id and entity kind.
sortKey: string, optional None
The key by which the entity list is sorted.
Possible values:
* `freqsort`: by frequency
* `sort_0` or `sort_eid`: by entity identifier
* `sort_1` or `sort_kind`: by entity kind
If `None` is passed, `freqsort` is filled in.
sortDir: string, optional None
The direction of the sort.
Possible values:
* `a`: ascending
* `d`: descending
If `None` is passed, `a` is filled in.
However, if `None` is passed for both `sortKey` and `sortDir`,
a `d` is filled in.
As a consequence, the default sort order is by frequency, most
frequent on top.
cutOffFreq: integer, optional None
If passed, it is a lower limit on the frequency of the entities that
will be shown. Every entity with a lower frequency will be skipped.
Returns
-------
string or void
If called by the browser, it returns the HTML string.
Otherwise, it displays the HTML string in the output, assuming
it is a cell in a Jupyter Notebook.
"""
settings = self.settings
features = settings.features
browse = self.browse
setData = self.getSetData()
hasEnt = activeEntity is not None
entries = setData.entityIdent.items()
sortKeyMap = {feat: i for (i, feat) in enumerate(features)}
if sortKey is None and sortDir is None:
(sortKey, sortDir) = SORT_DEFAULT
else:
if sortKey is None:
sortKey = SORTKEY_DEFAULT
if sortDir is None:
sortDir = SORTDIR_DEFAULT
if sortKey == SORTKEY_DEFAULT:
entries = sorted(entries, key=lambda x: (len(x[1]), x[0]))
else:
if sortKey.startswith("sort_"):
index = sortKey[5:]
if index.isdecimal():
index = int(index)
if index >= len(features):
index = 0
else:
index = sortKeyMap.get(index, 0)
else:
index = 0
entries = sorted(entries, key=lambda x: (x[0][index], -len(x[1])))
if sortDir == SORTDIR_DESC:
entries = reversed(entries)
content = []
for vals, es in entries:
x = len(es)
if cutOffFreq is not None and x < cutOffFreq:
continue
identRep = "⊙".join(vals)
active = " queried " if hasEnt and vals == activeEntity else ""
content.append(
H.p(
H.span(f"{x} ", cls="stat"),
repIdent(features, vals, active=active),
cls=f"e {active}",
enm=identRep,
)
)
content = H.join(content)
if browse:
return content
dh(content)
def showTriggers(
self,
activeEntity=None,
activeTrigger=None,
sortKey=None,
sortDir=None,
subtleFilter=None,
zeroFilter=0,
):
"""Generates HTML for an expandable overview of the entities and their triggers.
Parameters
----------
activeEntity: tuple, optional None
The entity that must be highlighted.
activeTrigger: tuple, optional None
The entity that must be highlighted.
sortKey: string, optional None
The key by which the entity list is sorted.
Possible values:
* `freqsort`: by frequency
* `sort_0` or `sort_name`: by entity name
If `None` is passed, `freqsort` is filled in.
sortDir: string, optional None
The direction of the sort.
Possible values:
* `a`: ascending
* `d`: descending
If `None` is passed, `a` is filled in.
However, if `None` is passed for both `sortKey` and `sortDir`,
a `d` is filled in.
As a consequence, the default sort order is by frequency, most
frequent on top.
subtleFilter: boolean, optional None
Filters on the kind of sheets in which triggers occur.
If None: all sheets are considered.
If True: only context sheets are considered.
If False: only the main sheet is considered.
zeroFilter: integer, optional 0
Filters on triggers with zero hits.
If `0`: no filtering.
If `1` or `-1`: shows all entities that have at least one trigger
with zero hits.
If `1`, for those entities, all triggers will be shown, also the
one with hits.
If `-1`, for those entities only the triggers with zero hits will be shown.
Only shows triggers without hits.
Returns
-------
string or void
If called by the browser, it returns the HTML string.
Otherwise, it displays the HTML string in the output, assuming
it is a cell in a Jupyter Notebook.
"""
browse = self.browse
sheetData = self.getSheetData()
hitData = sheetData.hitData or {}
nameMap = sheetData.nameMap or {}
rowMap = sheetData.rowMap or {}
hasEnt = activeEntity is not None
hasTrig = activeTrigger is not None
def entryNames(data):
entries = []
for eidkind, triggers in data.items():
if eidkind not in nameMap:
# should not occur!
continue
name = nameMap[eidkind]
tOccs = sum(triggers.values())
subtle = any(t[1] != "" for t in triggers)
if (
subtleFilter is None
or subtleFilter
and subtle
or not subtleFilter
and not subtle
):
entries.append(
(
eidkind,
name,
tOccs,
subtle,
entryTriggers(triggers),
)
)
return entries
def entryTriggers(data):
entries = []
for trigger, nOccs in sorted(data.items(), key=lambda x: (-x[1], x[0])):
subtle = trigger[1] != ""
entries.append((trigger, nOccs, subtle))
return entries
entries = entryNames(hitData)
if sortKey is None and sortDir is None:
(sortKey, sortDir) = SORT_DEFAULT
else:
if sortKey is None:
sortKey = SORTKEY_DEFAULT
if sortDir is None:
sortDir = SORTDIR_DEFAULT
if sortKey == SORTKEY_DEFAULT:
entries = sorted(entries, key=lambda x: (x[2], toAscii(x[1]).lower()))
else:
entries = sorted(entries, key=lambda x: (toAscii(x[1]).lower(), -x[2]))
if sortDir == SORTDIR_DESC:
entries = reversed(entries)
def genNames(data):
content = []
for eidkind, name, tOccs, subtle, triggers in entries:
identRep = "⊙".join(eidkind)
hasTriggerNoOccs = any(x[1] == 0 for x in triggers)
hasNoOccs = tOccs == 0
if zeroFilter and not (hasNoOccs or hasTriggerNoOccs):
continue
occsCls = (
"nooccs" if hasNoOccs else ("warnoccs" if hasTriggerNoOccs else "")
)
subtleCls = "subtle" if subtle else ""
active = "queried" if hasEnt and eidkind == activeEntity else ""
content.append(
H.div(
H.div((H.span(f"{tOccs} ", cls="stat"))),
H.div(
(
H.div(name, cls=f"ntx {occsCls} {subtleCls}"),
H.div(genTriggers(triggers), cls="etrigger"),
)
),
cls=f"e {active}",
enm=identRep,
)
)
return content
def genTriggers(data):
content = []
for tInfo, nOccs, subtle in data:
(trigger, scope) = tInfo
hasNoOccs = nOccs == 0
if zeroFilter == -1 and not hasNoOccs:
continue
rows = rowMap.get(trigger, [])
r = "?" if len(rows) == 0 else ", ".join(str(x) for x in rows)
occsCls = "nooccs" if nOccs == 0 else ""
subtleCls = "subtle" if subtle else ""
scopeRep = f"{SET_SHEET}{scope}"
triggerRep = "⊙".join(str(x) for x in tInfo)
active = (
"tqueried" if hasTrig and (trigger, scope) == activeTrigger else ""
)
content.append(
H.div(
(
H.span(f"{nOccs} ", cls="stat"),
H.span(
(
H.span(f"{scopeRep} "),
H.code(trigger, cls="ttx", title=f"r{r}"),
),
cls=f"{occsCls} {subtleCls}",
),
),
cls=f"et {active}",
etr=triggerRep,
),
)
return content
content = "\n".join(genNames(hitData))
if browse:
return content
dh(content)
def showContent(
self,
buckets,
activeEntity=None,
activeTrigger=None,
excludedTokens=set(),
mayLimit=True,
start=None,
end=None,
withNodes=False,
):
"""Generates HTML for a given portion of the corpus.
The corpus text will be marked up with entities, the positions of
these entities are present in the input parameter `buckets`.
It is recommended to apply this function to the outcome of
`tf.ner.corpus.Corpus.filterContent`
!!! caution "Truncated"
Unless the user has selected an entity or forced a start and end
boundary to the list of buckets, the display may be truncated.
See the parameter `mayLimit` below.
Parameters
----------
buckets: iterable of tuple
A selection of buckets (chunks / paragraphs) of the corpus.
Each bucket is given as a tuple.
The exact form of this data structure is equal to what the
function `tf.ner.corpus.Corpus.filterContent`
returns.
activeEntity: tuple, optional None
The entity that must be highlighted.
activeTrigger: tuple, optional None
The trigger that must be highlighted.
excludedTokens: set, optional None
If passed, it is a set of tokens where a ❌ has been placed by the
user. They correspond to occurrences that have been deselected from
being subject to add / delete operations.
mayLimit: boolean, optional False
It is possible that the buckets make up the whole corpus.
Although we have optimised things in such a way that the browser can handle
a webpage with thousands of pages of material in it, such large pages
may compromise the performance.
If the bucket set is potentially very large, and the `start` and `end`
parameters are not both specified, we will truncate the list of buckets
to a smallish value (see `settings.LIMIT_BROWSER` and `settings.LIMIT_NB`).
However, when there is an `activeEntity`, we assume the buckets are those
containing that entity, and that it is a limited set anyway, and in that
case we do not truncate.
start: integer, optional None
If passed, start rendering the buckets at this position.
end: integer, optional None
If passed, stop rendering the buckets at this position.
withNodes: boolean, optional None
Shows the node in each token.
Returns
-------
string or void
If called by the browser, it returns the HTML string.
Otherwise, it displays the HTML string in the output, assuming
it is a cell in a Jupyter Notebook.
"""
settings = self.settings
bucketType = self.bucketType
features = settings.features
style = self.style
ltr = self.ltr
browse = self.browse
setData = self.getSetData()
setIsX = self.setIsX
setIsRo = self.setIsRo
setIsSrc = self.setIsSrc
afterv = self.getAfter()
sectionHead = self.sectionHead
entityIdent = setData.entityIdent
entitySlotIndex = setData.entitySlotIndex
hasEnt = activeEntity is not None
hasTrig = activeTrigger is not None
if setIsX:
sheetData = self.getSheetData()
triggerFromMatch = sheetData.triggerFromMatch
limited = mayLimit and not hasEnt and (start is None or end is None)
limit = LIMIT_BROWSER if browse else LIMIT_NB
content = []
nB = 0
nBshown = 0
for b, bTokens, matches, positions in buckets:
nB += 1
if start is not None and nB < start:
continue
if end is not None and nB > end:
break
if limited and nBshown > limit:
content.append(
H.div(
f"Showing only the first {limit} {bucketType}s of all "
f"{len(buckets)} ones.",
cls="report",
)
)
break
nBshown += 1
charPos = 0
if setIsRo:
if setIsSrc:
allMatches = set(chain.from_iterable(matches))
else:
allMatches = {}
for match in matches:
trigger = triggerFromMatch.get(match, None)
if trigger is not None:
for m in match:
allMatches[m] = trigger
else:
allMatches = set()
endMatches = set()
for match in matches:
allMatches |= set(match)
endMatches.add(match[-1])
headContent = H.span(
H.span(sectionHead(b), cls="bhl ltr", title="show context"),
cls=f"bh {ltr}",
node=b,
)
subContent = []
for t, w in bTokens:
info = entitySlotIndex.get(t, None)
inEntity = False
if info is not None:
inEntity = True
for item in sorted(
(x for x in info if x is not None), key=lambda z: z[1]
):
(status, lg, ident) = item
identRep = "⊙".join(ident)
if status:
active = (
" queried " if hasEnt and ident == activeEntity else ""
)
subContent.append(
H.span(
H.span(abs(lg), cls="lgb"),
repIdent(features, ident, active=active),
" ",
H.span(len(entityIdent[ident]), cls="n"),
cls="es",
enm=identRep,
)
)
after = afterv(t) or ""
lenW = len(w)
lenWa = len(w) + len(after)
foundSet = set(range(charPos, charPos + lenW)) & positions
found = len(foundSet) != 0
if found:
firstFound = min(foundSet) - charPos
lastFound = max(foundSet) - charPos
leading = "" if firstFound == charPos else w[0:firstFound]
trailing = "" if lastFound == lenW - 1 else w[lastFound + 1 :]
hit = w[firstFound : lastFound + 1]
wRep = H.join(leading, H.span(hit, cls="found"), trailing)
else:
wRep = w
queried = t in allMatches
tqueried = (
setIsX and hasTrig and allMatches.get(t, None) == activeTrigger
)
hlClasses = " queried " if queried else ""
hlClasses += " tqueried " if tqueried else ""
hlClasses += " ei " if inEntity else ""
hlClasses += f" {style} " if style else ""
hlClass = dict(cls=hlClasses) if hlClasses else {}
endQueried = (not setIsRo) and t in endMatches
excl = "x" if t in excludedTokens else "v"
nodeRep = H.span(str(t), cls="nd") if withNodes else ""
subContent.append(
H.join(
H.span(wRep + nodeRep, **hlClass, t=t),
H.span(te=t, st=excl) if endQueried else "",
after,
)
)
if info is not None:
for item in sorted(
(x for x in info if x is not None), key=lambda z: z[1]
):
(status, lg, ident) = item
if not status:
subContent.append(
H.span(H.span(abs(lg), cls="lge"), cls="ee")
)
charPos += lenWa
content.append(
H.div(headContent, H.nb, H.span(subContent, cls=ltr), cls=f"b {ltr}")
)
if browse:
return H.div(content, id="buckets", cls=f"buckets {ltr}")
dh(H.div(content, cls=f"buckets {ltr}"))Classes
class Show-
Expand source code Browse git
class Show: def showEntityOverview(self): """Generates HTML for an overview of the entities. The entity overview consists of a listing of the possible entity kinds with for each kind how many entities there are of that kind. Returns ------- string or void If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook. """ settings = self.settings keywordFeatures = settings.keywordFeatures browse = self.browse setData = self.getSetData() content = H.p( H.span( H.code(f"{len(es):>5}"), " x ", H.span(repSummary(keywordFeatures, fVals)), ) + H.br() for (fVals, es) in sorted( setData.entitySummary.items(), key=lambda x: (-len(x[1]), x[0]) ) ) if browse: return content dh(content) def showEntities( self, activeEntity=None, sortKey=None, sortDir=None, cutOffFreq=None ): """Generates HTML for a sorted list of the entities. The entity list consists of a table of entity identifiers, with the columns for the kind and frequency of the entities. There may be one active entity, and that one will be highlighted. Parameters ---------- activeEntity: tuple, optional None The entity that must be highlighted, given as entity id and entity kind. sortKey: string, optional None The key by which the entity list is sorted. Possible values: * `freqsort`: by frequency * `sort_0` or `sort_eid`: by entity identifier * `sort_1` or `sort_kind`: by entity kind If `None` is passed, `freqsort` is filled in. sortDir: string, optional None The direction of the sort. Possible values: * `a`: ascending * `d`: descending If `None` is passed, `a` is filled in. However, if `None` is passed for both `sortKey` and `sortDir`, a `d` is filled in. As a consequence, the default sort order is by frequency, most frequent on top. cutOffFreq: integer, optional None If passed, it is a lower limit on the frequency of the entities that will be shown. Every entity with a lower frequency will be skipped. Returns ------- string or void If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook. """ settings = self.settings features = settings.features browse = self.browse setData = self.getSetData() hasEnt = activeEntity is not None entries = setData.entityIdent.items() sortKeyMap = {feat: i for (i, feat) in enumerate(features)} if sortKey is None and sortDir is None: (sortKey, sortDir) = SORT_DEFAULT else: if sortKey is None: sortKey = SORTKEY_DEFAULT if sortDir is None: sortDir = SORTDIR_DEFAULT if sortKey == SORTKEY_DEFAULT: entries = sorted(entries, key=lambda x: (len(x[1]), x[0])) else: if sortKey.startswith("sort_"): index = sortKey[5:] if index.isdecimal(): index = int(index) if index >= len(features): index = 0 else: index = sortKeyMap.get(index, 0) else: index = 0 entries = sorted(entries, key=lambda x: (x[0][index], -len(x[1]))) if sortDir == SORTDIR_DESC: entries = reversed(entries) content = [] for vals, es in entries: x = len(es) if cutOffFreq is not None and x < cutOffFreq: continue identRep = "⊙".join(vals) active = " queried " if hasEnt and vals == activeEntity else "" content.append( H.p( H.span(f"{x} ", cls="stat"), repIdent(features, vals, active=active), cls=f"e {active}", enm=identRep, ) ) content = H.join(content) if browse: return content dh(content) def showTriggers( self, activeEntity=None, activeTrigger=None, sortKey=None, sortDir=None, subtleFilter=None, zeroFilter=0, ): """Generates HTML for an expandable overview of the entities and their triggers. Parameters ---------- activeEntity: tuple, optional None The entity that must be highlighted. activeTrigger: tuple, optional None The entity that must be highlighted. sortKey: string, optional None The key by which the entity list is sorted. Possible values: * `freqsort`: by frequency * `sort_0` or `sort_name`: by entity name If `None` is passed, `freqsort` is filled in. sortDir: string, optional None The direction of the sort. Possible values: * `a`: ascending * `d`: descending If `None` is passed, `a` is filled in. However, if `None` is passed for both `sortKey` and `sortDir`, a `d` is filled in. As a consequence, the default sort order is by frequency, most frequent on top. subtleFilter: boolean, optional None Filters on the kind of sheets in which triggers occur. If None: all sheets are considered. If True: only context sheets are considered. If False: only the main sheet is considered. zeroFilter: integer, optional 0 Filters on triggers with zero hits. If `0`: no filtering. If `1` or `-1`: shows all entities that have at least one trigger with zero hits. If `1`, for those entities, all triggers will be shown, also the one with hits. If `-1`, for those entities only the triggers with zero hits will be shown. Only shows triggers without hits. Returns ------- string or void If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook. """ browse = self.browse sheetData = self.getSheetData() hitData = sheetData.hitData or {} nameMap = sheetData.nameMap or {} rowMap = sheetData.rowMap or {} hasEnt = activeEntity is not None hasTrig = activeTrigger is not None def entryNames(data): entries = [] for eidkind, triggers in data.items(): if eidkind not in nameMap: # should not occur! continue name = nameMap[eidkind] tOccs = sum(triggers.values()) subtle = any(t[1] != "" for t in triggers) if ( subtleFilter is None or subtleFilter and subtle or not subtleFilter and not subtle ): entries.append( ( eidkind, name, tOccs, subtle, entryTriggers(triggers), ) ) return entries def entryTriggers(data): entries = [] for trigger, nOccs in sorted(data.items(), key=lambda x: (-x[1], x[0])): subtle = trigger[1] != "" entries.append((trigger, nOccs, subtle)) return entries entries = entryNames(hitData) if sortKey is None and sortDir is None: (sortKey, sortDir) = SORT_DEFAULT else: if sortKey is None: sortKey = SORTKEY_DEFAULT if sortDir is None: sortDir = SORTDIR_DEFAULT if sortKey == SORTKEY_DEFAULT: entries = sorted(entries, key=lambda x: (x[2], toAscii(x[1]).lower())) else: entries = sorted(entries, key=lambda x: (toAscii(x[1]).lower(), -x[2])) if sortDir == SORTDIR_DESC: entries = reversed(entries) def genNames(data): content = [] for eidkind, name, tOccs, subtle, triggers in entries: identRep = "⊙".join(eidkind) hasTriggerNoOccs = any(x[1] == 0 for x in triggers) hasNoOccs = tOccs == 0 if zeroFilter and not (hasNoOccs or hasTriggerNoOccs): continue occsCls = ( "nooccs" if hasNoOccs else ("warnoccs" if hasTriggerNoOccs else "") ) subtleCls = "subtle" if subtle else "" active = "queried" if hasEnt and eidkind == activeEntity else "" content.append( H.div( H.div((H.span(f"{tOccs} ", cls="stat"))), H.div( ( H.div(name, cls=f"ntx {occsCls} {subtleCls}"), H.div(genTriggers(triggers), cls="etrigger"), ) ), cls=f"e {active}", enm=identRep, ) ) return content def genTriggers(data): content = [] for tInfo, nOccs, subtle in data: (trigger, scope) = tInfo hasNoOccs = nOccs == 0 if zeroFilter == -1 and not hasNoOccs: continue rows = rowMap.get(trigger, []) r = "?" if len(rows) == 0 else ", ".join(str(x) for x in rows) occsCls = "nooccs" if nOccs == 0 else "" subtleCls = "subtle" if subtle else "" scopeRep = f"{SET_SHEET}{scope}" triggerRep = "⊙".join(str(x) for x in tInfo) active = ( "tqueried" if hasTrig and (trigger, scope) == activeTrigger else "" ) content.append( H.div( ( H.span(f"{nOccs} ", cls="stat"), H.span( ( H.span(f"{scopeRep} "), H.code(trigger, cls="ttx", title=f"r{r}"), ), cls=f"{occsCls} {subtleCls}", ), ), cls=f"et {active}", etr=triggerRep, ), ) return content content = "\n".join(genNames(hitData)) if browse: return content dh(content) def showContent( self, buckets, activeEntity=None, activeTrigger=None, excludedTokens=set(), mayLimit=True, start=None, end=None, withNodes=False, ): """Generates HTML for a given portion of the corpus. The corpus text will be marked up with entities, the positions of these entities are present in the input parameter `buckets`. It is recommended to apply this function to the outcome of `tf.ner.corpus.Corpus.filterContent` !!! caution "Truncated" Unless the user has selected an entity or forced a start and end boundary to the list of buckets, the display may be truncated. See the parameter `mayLimit` below. Parameters ---------- buckets: iterable of tuple A selection of buckets (chunks / paragraphs) of the corpus. Each bucket is given as a tuple. The exact form of this data structure is equal to what the function `tf.ner.corpus.Corpus.filterContent` returns. activeEntity: tuple, optional None The entity that must be highlighted. activeTrigger: tuple, optional None The trigger that must be highlighted. excludedTokens: set, optional None If passed, it is a set of tokens where a ❌ has been placed by the user. They correspond to occurrences that have been deselected from being subject to add / delete operations. mayLimit: boolean, optional False It is possible that the buckets make up the whole corpus. Although we have optimised things in such a way that the browser can handle a webpage with thousands of pages of material in it, such large pages may compromise the performance. If the bucket set is potentially very large, and the `start` and `end` parameters are not both specified, we will truncate the list of buckets to a smallish value (see `settings.LIMIT_BROWSER` and `settings.LIMIT_NB`). However, when there is an `activeEntity`, we assume the buckets are those containing that entity, and that it is a limited set anyway, and in that case we do not truncate. start: integer, optional None If passed, start rendering the buckets at this position. end: integer, optional None If passed, stop rendering the buckets at this position. withNodes: boolean, optional None Shows the node in each token. Returns ------- string or void If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook. """ settings = self.settings bucketType = self.bucketType features = settings.features style = self.style ltr = self.ltr browse = self.browse setData = self.getSetData() setIsX = self.setIsX setIsRo = self.setIsRo setIsSrc = self.setIsSrc afterv = self.getAfter() sectionHead = self.sectionHead entityIdent = setData.entityIdent entitySlotIndex = setData.entitySlotIndex hasEnt = activeEntity is not None hasTrig = activeTrigger is not None if setIsX: sheetData = self.getSheetData() triggerFromMatch = sheetData.triggerFromMatch limited = mayLimit and not hasEnt and (start is None or end is None) limit = LIMIT_BROWSER if browse else LIMIT_NB content = [] nB = 0 nBshown = 0 for b, bTokens, matches, positions in buckets: nB += 1 if start is not None and nB < start: continue if end is not None and nB > end: break if limited and nBshown > limit: content.append( H.div( f"Showing only the first {limit} {bucketType}s of all " f"{len(buckets)} ones.", cls="report", ) ) break nBshown += 1 charPos = 0 if setIsRo: if setIsSrc: allMatches = set(chain.from_iterable(matches)) else: allMatches = {} for match in matches: trigger = triggerFromMatch.get(match, None) if trigger is not None: for m in match: allMatches[m] = trigger else: allMatches = set() endMatches = set() for match in matches: allMatches |= set(match) endMatches.add(match[-1]) headContent = H.span( H.span(sectionHead(b), cls="bhl ltr", title="show context"), cls=f"bh {ltr}", node=b, ) subContent = [] for t, w in bTokens: info = entitySlotIndex.get(t, None) inEntity = False if info is not None: inEntity = True for item in sorted( (x for x in info if x is not None), key=lambda z: z[1] ): (status, lg, ident) = item identRep = "⊙".join(ident) if status: active = ( " queried " if hasEnt and ident == activeEntity else "" ) subContent.append( H.span( H.span(abs(lg), cls="lgb"), repIdent(features, ident, active=active), " ", H.span(len(entityIdent[ident]), cls="n"), cls="es", enm=identRep, ) ) after = afterv(t) or "" lenW = len(w) lenWa = len(w) + len(after) foundSet = set(range(charPos, charPos + lenW)) & positions found = len(foundSet) != 0 if found: firstFound = min(foundSet) - charPos lastFound = max(foundSet) - charPos leading = "" if firstFound == charPos else w[0:firstFound] trailing = "" if lastFound == lenW - 1 else w[lastFound + 1 :] hit = w[firstFound : lastFound + 1] wRep = H.join(leading, H.span(hit, cls="found"), trailing) else: wRep = w queried = t in allMatches tqueried = ( setIsX and hasTrig and allMatches.get(t, None) == activeTrigger ) hlClasses = " queried " if queried else "" hlClasses += " tqueried " if tqueried else "" hlClasses += " ei " if inEntity else "" hlClasses += f" {style} " if style else "" hlClass = dict(cls=hlClasses) if hlClasses else {} endQueried = (not setIsRo) and t in endMatches excl = "x" if t in excludedTokens else "v" nodeRep = H.span(str(t), cls="nd") if withNodes else "" subContent.append( H.join( H.span(wRep + nodeRep, **hlClass, t=t), H.span(te=t, st=excl) if endQueried else "", after, ) ) if info is not None: for item in sorted( (x for x in info if x is not None), key=lambda z: z[1] ): (status, lg, ident) = item if not status: subContent.append( H.span(H.span(abs(lg), cls="lge"), cls="ee") ) charPos += lenWa content.append( H.div(headContent, H.nb, H.span(subContent, cls=ltr), cls=f"b {ltr}") ) if browse: return H.div(content, id="buckets", cls=f"buckets {ltr}") dh(H.div(content, cls=f"buckets {ltr}"))Subclasses
Methods
def showContent(self, buckets, activeEntity=None, activeTrigger=None, excludedTokens=set(), mayLimit=True, start=None, end=None, withNodes=False)-
Generates HTML for a given portion of the corpus.
The corpus text will be marked up with entities, the positions of these entities are present in the input parameter
buckets.It is recommended to apply this function to the outcome of
Corpus.filterContent()Truncated
Unless the user has selected an entity or forced a start and end boundary to the list of buckets, the display may be truncated. See the parameter
mayLimitbelow.Parameters
buckets:iterableoftuple- A selection of buckets (chunks / paragraphs) of the corpus.
Each bucket is given as a tuple.
The exact form of this data structure is equal to what the
function
Corpus.filterContent()returns. activeEntity:tuple, optionalNone- The entity that must be highlighted.
activeTrigger:tuple, optionalNone- The trigger that must be highlighted.
excludedTokens:set, optionalNone- If passed, it is a set of tokens where a ❌ has been placed by the user. They correspond to occurrences that have been deselected from being subject to add / delete operations.
mayLimit:boolean, optionalFalse-
It is possible that the buckets make up the whole corpus. Although we have optimised things in such a way that the browser can handle a webpage with thousands of pages of material in it, such large pages may compromise the performance. If the bucket set is potentially very large, and the
startandendparameters are not both specified, we will truncate the list of buckets to a smallish value (seesettings.LIMIT_BROWSERandsettings.LIMIT_NB).However, when there is an
activeEntity, we assume the buckets are those containing that entity, and that it is a limited set anyway, and in that case we do not truncate. start:integer, optionalNone- If passed, start rendering the buckets at this position.
end:integer, optionalNone- If passed, stop rendering the buckets at this position.
withNodes:boolean, optionalNone- Shows the node in each token.
Returns
stringorvoid- If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook.
def showEntities(self, activeEntity=None, sortKey=None, sortDir=None, cutOffFreq=None)-
Generates HTML for a sorted list of the entities.
The entity list consists of a table of entity identifiers, with the columns for the kind and frequency of the entities.
There may be one active entity, and that one will be highlighted.
Parameters
activeEntity:tuple, optionalNone- The entity that must be highlighted, given as entity id and entity kind.
sortKey:string, optionalNone-
The key by which the entity list is sorted.
Possible values:
freqsort: by frequencysort_0orsort_eid: by entity identifiersort_1orsort_kind: by entity kind
If
Noneis passed,freqsortis filled in. sortDir:string, optionalNone-
The direction of the sort.
Possible values:
a: ascendingd: descending
If
Noneis passed,ais filled in. However, ifNoneis passed for bothsortKeyandsortDir, adis filled in.As a consequence, the default sort order is by frequency, most frequent on top.
cutOffFreq:integer, optionalNone- If passed, it is a lower limit on the frequency of the entities that will be shown. Every entity with a lower frequency will be skipped.
Returns
stringorvoid- If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook.
def showEntityOverview(self)-
Generates HTML for an overview of the entities.
The entity overview consists of a listing of the possible entity kinds with for each kind how many entities there are of that kind.
Returns
stringorvoid- If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook.
def showTriggers(self, activeEntity=None, activeTrigger=None, sortKey=None, sortDir=None, subtleFilter=None, zeroFilter=0)-
Generates HTML for an expandable overview of the entities and their triggers.
Parameters
activeEntity:tuple, optionalNone- The entity that must be highlighted.
activeTrigger:tuple, optionalNone- The entity that must be highlighted.
sortKey:string, optionalNone-
The key by which the entity list is sorted.
Possible values:
freqsort: by frequencysort_0orsort_name: by entity name
If
Noneis passed,freqsortis filled in. sortDir:string, optionalNone-
The direction of the sort.
Possible values:
a: ascendingd: descending
If
Noneis passed,ais filled in. However, ifNoneis passed for bothsortKeyandsortDir, adis filled in.As a consequence, the default sort order is by frequency, most frequent on top.
subtleFilter:boolean, optionalNone-
Filters on the kind of sheets in which triggers occur.
If None: all sheets are considered.
If True: only context sheets are considered.
If False: only the main sheet is considered.
zeroFilter:integer, optional0- Filters on triggers with zero hits.
If
0: no filtering. If1or-1: shows all entities that have at least one trigger with zero hits. If1, for those entities, all triggers will be shown, also the one with hits. If-1, for those entities only the triggers with zero hits will be shown. Only shows triggers without hits.
Returns
stringorvoid- If called by the browser, it returns the HTML string. Otherwise, it displays the HTML string in the output, assuming it is a cell in a Jupyter Notebook.