def extract_no_analyst(text: str, date: datetime.date):
# Castellum höjs sitt behåll (sälj), med riktkurs 165 kronor (200)
tokens = tokenize(text)
direction_idx = single_idx_in_list(tokens, text_to_direction.keys())
sitt_idx = single_idx_in_list(tokens, ['sitt'])
riktkurs_idx = single_idx_in_list(tokens, ['riktkurs'])
if not (direction_idx < sitt_idx < riktkurs_idx):
return None
signal = Signal.from_text(tokens[sitt_idx + 1])
prev_signal_candidate = tokens[sitt_idx + 2]
if is_parentheses_enclosed(prev_signal_candidate):
prev_signal = Signal.from_text(prev_signal_candidate[1:-1])
else:
prev_signal = Signal.UNKNOWN
forecast_price = to_float(tokens[riktkurs_idx + 1])
prev_forecast_price = None
if len(tokens) > riktkurs_idx + 3:
prev_forecast_price = to_float(tokens[riktkurs_idx + 3][1:-1])
forecast = model.Forecast(extractor='no_analyst',
raw=text,
date=date,
analyst=None,
change_direction=Direction.from_text(
tokens[direction_idx]),
company=' '.join(tokens[0:direction_idx]),
signal=signal,
prev_signal=prev_signal,
forecast_price=forecast_price,
prev_forecast_price=prev_forecast_price,
currency=currencies[tokens[riktkurs_idx + 2]])
return forecast
def extract_bn(in_str: str, date: datetime.date):
# Morgan Stanley sänker riktkursen för Lundin Energy till 245 kronor (325), upprepar jämvikt - BN
# Deutsche Bank höjer riktkursen för Boliden till 250 kronor från 235 kronor. Rekommendationen köp upprepas. Det framgår av ett marknadsbrev.
# Pareto Securities höjer riktkursen för investmentbolaget Kinnevik till 290 kronor från 262 kronor, enligt en ny analys.
tokens = tokenize(in_str)
direction_idx = idx_in_list(tokens, text_to_direction.keys())
for_idx = idx_in_list(tokens, ['för'])
till_idx = idx_in_list(tokens, ['till'])
fran_idx = idx_in_list(tokens, ['från'])
signal_idx = idx_in_list(tokens, text_to_signal.keys())
if len(direction_idx) == 0 or len(till_idx) == 0 or len(for_idx) == 0:
return None
if direction_idx[0] > for_idx[0] or for_idx[0] > till_idx[0]:
return None
signal = Signal.UNKNOWN
prev_signal = Signal.UNKNOWN
if len(signal_idx) > 0:
signal = Signal.from_text(tokens[signal_idx[0]])
if 'upprepar' in tokens or 'upprepas' in tokens:
prev_signal = signal
elif len(tokens) > signal_idx[0] + 1:
prev_signal_candidate = tokens[signal_idx[0] + 1]
if is_parentheses_enclosed(prev_signal_candidate):
prev_signal = Signal.from_text(prev_signal_candidate[1:-1])
forecast_price = to_float(tokens[till_idx[0] + 1])
currency = None
prev_forecast_price = None
if forecast_price is not None:
currency = currencies[tokens[till_idx[0] + 2]]
if len(fran_idx) != 0:
if tokens[fran_idx[0] + 1].isnumeric():
prev_forecast_price = to_float(tokens[fran_idx[0] + 1])
elif tokens[fran_idx[0] + 2].isnumeric():
prev_forecast_price = to_float(tokens[fran_idx[0] + 2])
else:
prev_forecast_price_candidate = tokens[till_idx[0] + 3]
if is_parentheses_enclosed(prev_forecast_price_candidate):
prev_forecast_price = to_float(
prev_forecast_price_candidate[1:-1])
forecast = model.Forecast(
extractor='bn',
raw=in_str,
date=date,
analyst=' '.join(tokens[:direction_idx[0]]),
change_direction=Direction.from_text(tokens[direction_idx[0]]),
company=' '.join(tokens[for_idx[0] + 1:till_idx[0]]),
signal=signal,
prev_signal=prev_signal,
forecast_price=forecast_price,
prev_forecast_price=prev_forecast_price,
currency=currency)
return forecast
def extract_simple(text: str, date: datetime.date):
# Kepler Cheuvreux sänker LVMH till behåll (köp), riktkurs 400 euro.
# Bank of America Merrill Lynch sänker EQT till underperform (neutral)
if 'riktkursen för' in text:
return None
tokens = tokenize(text)
direction_idx = idx_in_list(tokens, text_to_direction.keys())
till_idx = idx_in_list(tokens, ['till'])
signal_idx = idx_in_list(tokens, text_to_signal.keys())
forecast_idx = idx_in_list(tokens, ['riktkurs'])
if len(direction_idx) != 1 or len(till_idx) != 1 or len(
forecast_idx) > 1 or len(signal_idx) != 1:
return None
if direction_idx[0] > till_idx[0] or till_idx[0] > signal_idx[0]:
return None
prev_signal = Signal.UNKNOWN
if signal_idx[0] + 1 != len(tokens):
prev_signal_candidate = tokens[signal_idx[0] + 1]
if is_parentheses_enclosed(prev_signal_candidate):
prev_signal = Signal.from_text(prev_signal_candidate[1:-1])
forecast_price = None
prev_forecast_price = None
currency = None
if len(forecast_idx) == 1:
forecast_price = to_float(tokens[forecast_idx[0] + 1])
if forecast_price is not None:
if tokens[forecast_idx[0] - 1] == 'upprepar':
prev_forecast_price = forecast_price
elif len(tokens) > forecast_idx[0] + 3:
prev_forecast_price_candidate = tokens[forecast_idx[0] + 3]
if is_parentheses_enclosed(prev_forecast_price_candidate):
prev_forecast_price = to_float(
prev_forecast_price_candidate[1:-1])
currency = currencies[tokens[forecast_idx[0] + 2]]
forecast = model.Forecast(
extractor='simple',
raw=text,
date=date,
analyst=' '.join(tokens[:direction_idx[0]]),
change_direction=Direction.from_text(tokens[direction_idx[0]]),
company=' '.join(tokens[direction_idx[0] + 1:till_idx[0]]),
signal=Signal.from_text(tokens[signal_idx[0]]),
prev_signal=prev_signal,
forecast_price=forecast_price,
prev_forecast_price=prev_forecast_price,
currency=currency)
return forecast
def extract_forecast(text: str, date: datetime.date):
logging.debug(f"Extracting: {text}")
result = extract_simple(text, date)
if result is not None:
return result
result = extract_bloomberg(text, date)
if result is not None:
return result
result = extract_bn(text, date)
if result is not None:
return result
result = extract_no_analyst(text, date)
if result is not None:
return result
result = extract_inled(text, date)
if result is not None:
return result
result = extract_motivated_value(text, date)
if result is not None:
return result
return model.Forecast(raw=text, date=date)
def extract_motivated_value(text: str, date: datetime.date):
# "Redeye höjer motiverat värde för Systemair till 168 kronor (155)."
# Redeye höjer sitt motiverade värde i basscenariot för bettingbolaget Enlabs till 30 kronor, från tidigare 29 kronor.
# Might not be needed anymore after extract_bn generalised.
tokens = tokenize(text)
motiverat_idx = single_idx_in_list(tokens, ['motiverat'])
varde_idx = single_idx_in_list(tokens, ['värde'])
for_idx = single_idx_in_list(tokens, [
'för',
])
till_idx = single_idx_in_list(tokens, ['till'])
if motiverat_idx + 1 != varde_idx:
return None
if for_idx > till_idx:
return None
prev_forecast_price = None
if len(tokens) - 1 >= till_idx + 3 and is_parentheses_enclosed(
tokens[till_idx + 3]):
prev_forecast_price = to_float(tokens[till_idx + 3][1:-1])
currency = currencies[tokens[till_idx + 2]]
forecast = model.Forecast(extractor='motivated_value',
raw=text,
date=date,
analyst=' '.join(tokens[:motiverat_idx - 1]),
change_direction=Direction.from_text(
tokens[motiverat_idx - 1]),
company=' '.join(tokens[for_idx + 1:till_idx]),
signal=Signal.UNKNOWN,
prev_signal=Signal.UNKNOWN,
forecast_price=to_float(tokens[till_idx + 1]),
prev_forecast_price=prev_forecast_price,
currency=currency)
return forecast
def extract_bloomberg(text: str, date: datetime.date):
# "Goldman Sachs & Co sänker sin rekommendation för Outokumpu till neutral från köp."
tokens = tokenize(text)
sin_idx = single_idx_in_list(tokens, ['sin'])
for_idx = single_idx_in_list(tokens, ['för'])
till_idx = single_idx_in_list(tokens, ['till'])
fran_idx = single_idx_in_list(tokens, ['från'])
if not (0 < sin_idx < for_idx < till_idx < fran_idx):
return None
forecast = model.Forecast(
extractor='bloomberg',
raw=text,
date=date,
analyst=' '.join(tokens[:sin_idx - 1]),
change_direction=Direction.from_text(tokens[sin_idx - 1]),
company=' '.join(tokens[for_idx + 1:till_idx]),
signal=Signal.from_text(tokens[till_idx + 1]),
prev_signal=Signal.from_text(tokens[fran_idx + 1]),
forecast_price=None,
prev_forecast_price=None,
currency=None)
return forecast
def extract_inled(text: str, date: datetime.date):
# BTIG inleder bevakning på Tripadvisor med rekommendationen neutral.
tokens = tokenize(text)
inleder_idx = idx_in_list(tokens, ['inleder'])
med_idx = idx_in_list(tokens, ['med'])
if len(inleder_idx) != 1 or len(med_idx) != 1:
return None
if inleder_idx[0] > med_idx[0]:
return None
forecast = model.Forecast(extractor='inled',
raw=text,
date=date,
analyst=' '.join(tokens[:inleder_idx[0]]),
change_direction=Direction.NEW,
company=' '.join(tokens[inleder_idx[0] +
3:med_idx[0]]),
signal=Signal.from_text(tokens[med_idx[0] + 2]),
prev_signal=Signal.UNKNOWN,
forecast_price=None,
prev_forecast_price=None,
currency=None)
return forecast