Zviri Mukati[Viga][Ratidza]
Mazwi ekufona ari kubviswa achifarira zvinyorwa uye zvinoonekwa muchikamu chekutaurirana. Zvinoenderana nesarudzo yeFacebook, vanopfuura hafu yevatengi vanofarira kutenga kubva kukambani yavanogona kutaura nayo. Kukurukurirana kwazova nzira itsva inogamuchirika munzanga.
Inogonesa mabhizinesi kutaurirana nevatengi vavo chero nguva uye kubva chero nzvimbo. Chatbots iri kuwedzera mukurumbira pakati pemakambani nevatengi nekuda kwekureruka kwekushandisa uye kuderedzwa kwekumirira nguva.
Chatbots, kana otomatiki ekukurukurirana zvirongwa, inopa vatengi imwe yakasarudzika nzira yekuwana masevhisi kuburikidza nemavara-based interface. Iwo matsva eAI-powered chatbots anogona kuziva mubvunzo (mubvunzo, kuraira, kurongeka, nezvimwewo) wakaitwa nemunhu (kana imwe bot, kutanga) mune yakatarwa nharaunda uye kupindura nenzira kwayo (mhinduro, chiito, nezvimwewo).
Mune ino post, isu tichaenda pamusoro pekuti ma chatbots chii, mabhenefiti avo, makesi ekushandisa, uye maitiro ekugadzira yako kudzidza zvakadzika chatbot muPython, pakati pezvimwe zvinhu.
Ngatitangei.
Saka, chii chinonzi chatbots?
Chatbot inowanzonzi ndeimwe yemhando yepamusoro uye inovimbisa nzira dzekudyidzana kwevanhu-muchina. Aya madhijitari vabatsiri anovandudza ruzivo rwevatengi nekugadzirisa kudyidzana pakati pevanhu nemasevhisi.
Saizvozvowo, vanopa mabhizinesi nesarudzo nyowani dzekukwirisa mutengi wekuonana nekuita basa, izvo zvinogona kuderedza zvakajairwa mari yekutsigira.
Muchidimbu, iAI-based software inorehwa kutaurirana nevanhu mumitauro yavo yechisikigo. Aya ma chatbots anowanzo pindirana kuburikidza neaudio kana kunyora matekiniki, uye anogona kutevedzera zviri nyore mitauro yevanhu kuitira kuti abatanidze nevanhu senge munhu.
Chatbots vanodzidza kubva mukudyidzana kwavo nevashandisi, vachiwedzera kuona uye kushanda nesimba nekufamba kwenguva. Ivo vanogona kubata rakasiyana siyana rezviitwa zvebhizinesi, sekubvumidza kushandisa mari, kuita nevatengi online, uye kugadzira inotungamira.
Kugadzira yako yakadzika yekudzidza chatbot nepython
Kune akawanda akasiyana emhando dzechatbots mumunda we machine learning uye AI. Mamwe ma chatbots anonyatso vabatsiri, nepo vamwe varipo kuti vataure navo, nepo vamwe vari vamiririri vebasa revatengi.
Iwe wakamboona vamwe vanoshandirwa nemabhizinesi kupindura mibvunzo. Tichagadzira diki chatbot muchidzidzo chino kupindura mibvunzo inowanzo bvunzwa.
1. Kuisa mapakeji
Nhanho yedu yekutanga ndeyekuisa anotevera mapakeji.
2. Kudzidzisa Data
Ikozvino yave nguva yekufunga kuti ndeupi rudzi rweruzivo rwatinozoda kupa yedu chatbot. Hatidi kudhawunirodha chero akakura madataset nekuti iyi iri nyore chatbot.
Tichangoshandisa ruzivo rwatakagadzira isu pachedu. Kuti unyatsotevera chidzidzo, unozofanirwa kugadzira .JSON faira rine mafomati akafanana neanoonekwa pazasi. Faira rangu rakanzi "intents.json."
Iyo JSON faira inoshandiswa kugadzira seti yemameseji ayo mushandisi angangopinza uye mepu kune seti yemhinduro dzakakodzera. Duramazwi rega rega riri mufaira rine tag inoratidza kuti meseji yega yega ndeyeboka ripi.
Tichashandisa ruzivo urwu kudzidzisa a neural network kuisa mutsara wemazwi seimwe yema tag mufaira redu.
Tinogona kuzongotora mhinduro kubva kumapoka iwayo toipa kumushandisi. Iyo chatbot ichave iri nani uye yakanyanya kuomarara kana iwe ukaipa nemamwe ma tag, mhinduro, uye mapatani.
3. JSON data kurodha
Tichatanga nekuisa mune yedu .json data uye kupinza mamwe ma module. Unganidza your.json faira mudhairekitori rakafanana nerako Python script. Yedu .json data ikozvino ichachengetwa mune data variable.
4. Kutora Data
Ikozvino yave nguva yekuburitsa ruzivo rwatinoda kubva kune yedu JSON faira. Yese mapatani, pamwe nekirasi / tag yavari, inodiwa.
Tichadawo rondedzero yeakasarudzika mazwi mumaitiro edu (nekuda kwezvikonzero zvatichatsanangura gare gare), saka ngatigadzire zvinyorwa zvisina kunyorwa kuti tirambe tichitevera izvi zvakakosha.
Iye zvino tichapenengura data redu reJSON uye totora ruzivo rwatinoda. Pane kuti tiite setambo, tichashandisa nltk.word tokenizer kushandura pateni yega yega kuita rondedzero yemazwi.
Zvino, mune yedu docs_x runyorwa, isu tichawedzera imwe neimwe pateni, pamwe neinobatanidzwa tag, kune iyo docs_y runyorwa.
5. Kutsika kweShoko
Kuwana mudzi wezwi kunozivikanwa se stemming. Somuenzaniso, dzinde reshoko rokuti “icho” dzinde ringava “izvo,” nepo dzinde reshoko rokuti “zvinoitika” richigona kuva “zvinoitika.”
Isu tichashandisa iyi stiming nzira kudzikisa pasi izwi remodhi yedu uye kuyedza kufunga kuti mitsara inorevei mune zvese. Iyi kodhi inongoburitsa yakasarudzika runyorwa rwemashoko akatsikiswa anozoshandiswa muchikamu chinotevera chegadziriro yedu yedata.
6. Homwe Yemashoko
Yave nguva yekutaura nezve bhegi remashoko izvozvi zvataunza kunze data redu uye kugadzira mazwi akadzikama. Neural network uye maalgorithms ekudzidza muchina, sezvatinoziva tese, zvinoda kuiswa kwenhamba. Saka tambo yedu runyoro haisi kuzoigura. Tinoda nzira yekumiririra nhamba mumitsara yedu, uko ndiko kunopinda bhegi remashoko.
Chitsauko chega chega chichamiririrwa nerondedzero yehurefu hwenhamba yematemu mumutauro wemodhi yedu. Izwi rega rega muduramazwi redu richamiririrwa nenzvimbo iri muhurongwa. Kana chinzvimbo chiri muchirongwa chiri 1, izwi rinowanikwa mukutaura kwedu; kana ari 0, izwi harioneki mumutsara wedu.
Tinoridaidza kuti bhegi remazwi nekuti hatizive kutevedzana kwemazwi mumutsara; chatinongoziva ndechekuti dziripo mumutauro wemuenzaniso wedu.
Pamusoro pekugadzirisa mapindiro edu, isu tinofanirawo kufomatidza zvatinobuda kuitira kuti neural network inzwisise. Tichavaka zvinyorwa zvinobuda zvinova kureba kwenhamba yemavara/matagi mudataset yedu, zvakafanana nebhegi remashoko. Nzvimbo yega yega murunyorwa inomiririra yakasarudzika label/tag, uye 1 mune chero ipi yenzvimbo idzodzo inoratidza kuti nderipi label/tag iri kumiririrwa.
Chekupedzisira, isu tichashandisa NumPy arrays kuchengetedza yedu yekudzidziswa data uye zvinobuda.
7. Model Development
Isu takagadzirira kutanga kuvaka nekudzidzisa modhi ikozvino nekuti isu takagadziridza ese data redu. Isu tichashandisa yakakosha feed-forward neural network ine maviri akavanzika akaturikidzana ezvinangwa zvedu.
Chinangwa chetiweki yedu chichava chekutarisa kuunganidzwa kwemazwi uye kugovera kukirasi (imwe yema tag edu kubva kuJSON file). Isu tinotanga nekumisikidza yedu yemhando yekuvaka. Ramba uchifunga kuti iwe unogona kutamba nedzimwe nhamba kuti uuye nemhando iri nani! Machine pakudzidza kazhinji yakavakirwa pakuedza uye kukanganisa.
8. Muenzaniso Kudzidzisa & Kuchengetedza
Yave nguva yekudzidzisa modhi yedu pane yedu data ikozvino isu tazvimisa! Isu tichazadzisa izvi nekukodzera data redu kune modhi. Huwandu hwema epoch atinopa ndiyo nhamba yenguva iyo modhi icharatidzwa kune imwechete data panguva yekudzidziswa.
Tinogona kuchengetedza modhi kune iyo faira modhi kana tapedza kuidzidzisa. tflearn chinyorwa chinogona kushandiswa mune mamwe mascript.
9. Kushandisa chatbot
Iye zvino unogona kutanga kutaura ne bot yako.
Mabhenefiti eChatbot
- Sezvo mabhoti anotarisirwa kushanda mazuva 365 pagore, maawa makumi maviri nemana pazuva, pasina mubhadharo, kuwedzera kuwanikwa uye kukurumidza kuita.
- Aya mabhoti maturusi akakwana ekubata hombe data matatu makiyi Vs: vhoriyamu, velocity, uye akasiyana.
- Chatbots isoftware inogona kushandiswa kudzidza nezve uye kunzwisisa vatengi vekambani.
- Iyo ine simba repamusoro zvekuti ine mutengo wakachipa wekugadzirisa mushure mekuva nemabhenefiti epamusoro.
- Chatbot Zvishandiso zvinogadzira data rinogona kuchengetedzwa uye kushandiswa pakuongorora uye kufanotaura.
Usecase
- Kugadzirisa mibvunzo yemutengi
- Kupindura mibvunzo inowanzo bvunzwa
- Kugovera vatengi kutsigira timu
- Kuunganidza mhinduro dzevatengi
- Kukurudzira zvitsva zvinopihwa
- Shop with conversational commerce
- IT Helpdesk
- Kubhuka pekugara
- Kuendeswa kwemari
mhedziso
Chatbots, semamwe matekinoroji eAI, achashandiswa kuwedzera hunyanzvi hwevanhu uye kusunungura vanhu kuti vawedzere kugadzira uye kufungidzira nekuvabvumira kushandisa nguva yakawanda pane zvehunyanzvi kwete mabasa ehungwaru.
Mabhizinesi, vashandi, uye vatengi vanogona kubatsirwa kubva kune akakwenenzverwa chatbot maficha akadai nekukurumidza kurudziro uye fungidziro, pamwe nekuwana nyore kune yakakwirira-tsananguro yevhidhiyo musangano kubva mukati mehurukuro, munguva pfupi iri kutevera, apo AI inosanganiswa nekuvandudza 5G tekinoroji.
Izvi nezvimwe zvingangoitika zvichiri kuferefetwa, asi sekubatana kweinternet, AI, NLP, uye kufambira mberi kwekudzidza kwemuchina, ivo vanozonyanya kuwanda.
Chwoo
Mhoro,
Ndatenda nechirongwa ichi.
Ndine mubvunzo.
“Bag_of_words” harina kutsanangurwa. Handisi kunzwisisa kukanganisa uku.
Mungandiudzewo kuti ndingagadzirisa sei kukanganisa uku??
Ndatenda nechirongwa ichi!! Iva nezuva rakanaka
Jay
Ndokumbira uwedzere basa usati washandisa chatbot chikamu:
///////////////////////////////////// ////////////////////////
def bag_of_words(s, mazwi):
bhegi = [0 ye _ mukati (len (mashoko))]
s_words = nltk.word_tokenize(s)
s_words = [stemmer.stem(word.lower()) for word in s_words]
for se in s_words:
nokuti ini, w mukuverenga(mashoko):
kana w == se:
bhegi[i] = 1
return numpy.array(bag)
// Zvichagadzirisa nyaya yako. //
////////////////////////////////////// // ///////////////////////
Ndiri kugovera kodhi yakazara newe, saka uchawana mufananidzo wakajeka wayo.
///////////////////////////////////// /////////
pinza nltk
kubva nltk.stem.lancaster import LancasterStemmer
stemmer = LancasterStemmer ()
import numpy
import tflearn
import tensorflow
ekunze zvisingaite
kupinza json
kupinza pickle
ne open("intents.json") sefaira:
data = json.load(file)
edza:
with open(“data.pickle”, “rb”) as f:
mazwi, mavara, kudzidziswa, kubuda = pickle.load(f)
kunze kwe:
mazwi = []
mavara = []
docs_x = []
docs_y = []
nechinangwa che data[“zvinangwa”]:
yepateni muchinangwa["mienzaniso"]:
wrds = nltk.word_tokenize(pattern)
word.extend(wrds)
docs_x.append(wrds)
docs_y.append(chinangwa[“tag”])
kana chinangwa["tag"] chisiri mumazita:
labels.append(chinangwa[“tag”])
mazwi = [stemmer.stem(w.lower()) ye w mumashoko kana w != “?”]
mazwi = akarongwa(rondedzero(seti(mashoko)))
mavara = akarongwa (mavara)
kudzidzisa = []
zvabuda = []
out_empty = [0 for _ in range(len(labels))]
ye x, doc mune enumerate(docs_x):
bhegi = []
wrds = [stemmer.stem(w.lower()) for w in doc]
nokuti w mumashoko:
kana w mumashoko:
bag.append(1)
zvimwe:
bag.append(0)
output_row = out_empty[:]
output_row[labels.index(docs_y[x])] = 1
training.append(bhegi)
output.append(output_row)
training = numpy.array(training)
zvabuda = numpy.array(zvakabuda)
with open(“data.pickle”, “wb”) as f:
pickle.dump((mashoko, mavara, kudzidziswa, kubuda), f)
tensorflow.reset_default_graph()
net = tflearn.input_data(shape=[Hapana, len(training[0])])
net = tflearn.fully_connected(net, 8)
net = tflearn.fully_connected(net, 8)
net = tflearn.fully_connected(mambure, len(zvakabuda[0]), activation=”softmax”)
net = tflearn.regression(net)
model = tflearn.DNN(net)
edza:
model.load("model.tflearn")
kunze kwe:
model.fit(kudzidziswa, kubuda, n_epoch=1500, batch_size=8, show_metric=Chokwadi)
model.save(“model.tflearn”)
def bag_of_words(s, mazwi):
bhegi = [0 ye _ mukati (len (mashoko))]
s_words = nltk.word_tokenize(s)
s_words = [stemmer.stem(word.lower()) for word in s_words]
for se in s_words:
nokuti ini, w mukuverenga(mashoko):
kana w == se:
bhegi[i] = 1
return numpy.array(bag)
def chat():
print("Tanga kutaura ne bot (type quit to stop)!")
neChokwadi:
inp = kuisa("Iwe:")
kana inp.lower() == "siya":
zororo
mhinduro = model.predict([bag_of_words(inp, words)])
results_index = numpy.argmax(zvawanikwa)
tag = mavara[results_index]
ye tg mune data[“zvinangwa”]:
kana tg['tag'] == tag:
mhinduro = tg['mhinduro']
dhinda(random.choice(mhinduro))
chat()
///////////////////////////////////// //////////////
Ndatenda,
Anofara kukodha!
Lu
Mhoro,
Unogona here kundipa pfungwa yemaitiro ekuita munyaya yekuda kugadzira chatbot mupython, asi ruzivo rwunowanikwa kubva muongororo yeexcel. Ndatenda!