202 to resolve an issue with the electronic device 202. In one example, the troubleshooting solution may provide a series of troubleshooting steps that allow a user to overcome an issue related to the electronic device 202. In one example, the troubleshooting solution may be directly provided to the user. In another example, the troubleshooting solution may be communicated to a human support agent which in turn may provide the troubleshooting solution to the user of the electronic device 202.

In one example, the machine learning engine 106, running the machine learning model of the system 102, may be initially trained to identify the troubleshooting solutions for different issues associated with different functional features of various device models of the electronic device. For example, the system 102 may be trained to predict troubleshooting solutions for different issues associated with different functional features of different device models of imaging devices, such as printers and scanners. In another example, the system 102 may be trained to predict troubleshooting solutions for issues associated with functional features of multiple device models of networking devices, such as routers.

In one example, the system 102 may be trained to identify and provide the troubleshooting solutions for different issues associated with different functional features irrespective of the device model of the electronic device. To train the system 102, the representation generation engine 216 may initially analyze a plurality of troubleshoot support documents currently used for providing troubleshooting solution for the electronic devices. In one example, the troubleshoot support manual may include a plurality of pages having the troubleshoot support documents for various device models. In one example, a troubleshoot support document may refer to a page, one or more pages, logs, records, and parts or sections of a page. For instance, a single page of the troubleshoot support manual may include a troubleshoot support document that may provide troubleshooting steps to overcome an issue. In another example, a troubleshoot support document may span multiple pages of the troubleshoot support manual that may provide troubleshooting steps for an issue.

In one example, a troubleshoot support document may include a series of instructions or steps to resolve an issue related to a particular functionality of an electronic device. For instance, a first troubleshoot support document for imaging devices may include troubleshooting solution for resolving a paper scan issue. Similarly, a second troubleshoot support document for imaging devices may include troubleshooting solution for resolving a color mismatch issue. Similarly, a first troubleshoot support document for networking devices may include troubleshooting solution for resolving a network connection issue. Thus, for each combination of a functional feature and an issue with the feature a separate troubleshoot support document may be created, for example, by a manufacturer of the electronic device.

Further, each troubleshoot support document may indicate the device models of the electronic device 202 for which the troubleshoot support document may be used. In one example, the troubleshoot support document for a particular issue of a particular functional feature may indicate the device models that have the particular functional feature. For instance, the first troubleshoot support document for imaging devices may indicate all device models having a functional feature of scanning. The second troubleshoot support document for imaging devices may indicate all device models having a functional feature of colored printing, colored scanning, and colored copying.

In one example, the representation generation engine 216 may analyze the plurality of troubleshoot support documents to generate a high dimension support vector for each of the plurality of device models. The high dimension support vector for a device model may indicate a relationship between the device model and each of the plurality of support pages. In one example, the relationship may indicate either presence or absence of a reference of the device model on the troubleshoot support documents. An example illustration of support vectors for a plurality of electronic devices has been provided below. For the purpose of illustration, the high dimension support vectors have been shown for only eight troubleshoot support documents, however, each high dimension support vector may include values for a multiple troubleshoot support documents.

Example High Dimension Support Vectors

Trouble
Trouble
Trouble
Trouble
Trouble
Trouble
Trouble
Trouble

shoot
shoot
shoot
shoot
shoot
shoot
shoot
shoot

support
support
support
support
support
support
support
support

document
document
document
document
document
document
document
document

1
2
3
4
5
6
7
8

Device 1
1
0
1
1
0
0
1
0

Device 2
0
1
0
0
1
0
1
1

Device 3
1
0
0
0
0
1
0
0

As illustrated above, the high dimension support vectors for device models device 1, device 2, and device 3 are represented having values ‘0’ and ‘1’. In one example, value ‘1’ may indicate presence of a reference of the device model for a particular troubleshoot support document while the value ‘0’ may indicate absence of the reference of the device for that particular troubleshoot support document. In another example, value ‘1’ may indicate presence of a reference of the device model for a particular troubleshoot support document while the value ‘0’ may indicate absence of the reference of the device for that particular troubleshoot support document. For instance, in the above illustrated example, the device model, device 1 is understood to have been mentioned in four troubleshoot support documents—troubleshoot support documents 1, 3, 4, and 7.

Subsequently, the representation generation engine 216 may convert the high dimension support vectors to corresponding support vectors of lower dimension using a dimensionality reduction technique. Examples of the dimensionality reduction technique may include, but are not limited to, principal component analysis (PCA), kernel PCA, linear discriminant analysis (LDA), generalized discriminant analysis (GDA), and autoencoder. In one example, principal component analysis may be applied to the high dimension support vector to convert the high dimension support vector to a support vector of lower dimension. For example, high dimension support vectors for device models device 1, device 2, and device 3 may be converted to support vectors of lower dimension using principal component analysis. An example illustration of support vector of lower dimension is provided below. For the purpose of illustration, the support vector of lower dimension has been shown for only eight troubleshoot support documents, however, each low dimension support vector may include values for a multiple troubleshoot support documents.

Example Support Vectors of Low Dimension

Device 1
0.1
0.5
0.4

Device 2
0.5
0.2
0.3

Device 3
0.7
0.1
0.2

As illustrated, the support vector of lower dimension may include lesser values in comparison to the high dimension support vector. Thus, using support vectors of lower dimension may help in reducing complexity and resource utilization for using the machine learning model.

Further, the representation generation engine 216 may analyze the plurality of troubleshoot support documents using natural language processing to obtain a device feature vector for each device model. The device feature vector may indicate functional features of the device models. In one example, functional features may represent behavioral objectives associated with the particular device model. For example, the functional features of an imaging device may include scanning, fax, printing, and copying. Similarly, the functional features of a network peripheral may include routing and repeating messages. An example illustration of device feature vector for device models, device 1, device 2, and device 3 is provided below. For the purpose of illustration, the device feature vector has been shown for only three device features, however, each device model may support multiple device features.

Example Device Feature Vectors

ADF (Automatic

Scanner
Fax
Document Feature)

Device 1
1
1
0

Device 2
0
1
1

Device 3
1
1
1

As illustrated above, the device feature vectors for the device models device 1, device 2, and device 3 may include values ‘0’ and ‘1’ to indicate absence or presence of a functional feature. In one example, value ‘0’ may represent absence of a functional feature for the respective device model while value ‘1’ may represent presence of a functional feature for the respective device model. In another example, value ‘1’ may represent absence of a functional feature for the respective device model while value ‘0’ may represent presence of a functional feature for the respective device model. For instance, in the above illustrated example, the device 3 includes the scanning and fax functionality while the ADF functionality is not present.

The representation generation engine 216 may generate a context vector for each device model based on the support vector and device feature vector corresponding to the device model. In one example, the context vector for a device model may be generated by concatenating the device feature vector and the low dimensional support vector corresponding to the device model. Thus, corresponding context vector may be generated for each of the plurality of device models to obtain a plurality of context vectors. The plurality of context vectors corresponding to the plurality of device models may be saved in the representation generation data 224 and provided to the machine learning engine 106. The machine learning engine 106 may subsequently be trained to identify and predict the troubleshooting solution for different issues of the multiple device models of the electronic device using the feature vector, the issue, and the context vectors.

Further, the machine learning engine 106 may generate a mapping table including a mapping between the plurality of context vectors and the plurality of electronic device models. In one example, the mapping table may indicate, for each device model, the context vector corresponding to device model. The mapping table may be stored in the machine learning data 222 and may be updated each time a new device is added to the system 102. In case, a new device model is launched for which troubleshooting solutions are to be provided, the representation generation engine 216 may generate a new high dimension support vector for the new device model indicating a relationship between the new device model and each of the plurality of troubleshoot support documents. Further, a new support vector of lower dimension may be generated using the dimensionality reduction techniques as previously described.

Further, the representation generation engine 216 may generate a new device feature vector indicating the functional feature of the new device model. Subsequently, a new context vector may be generated for the new device model based on the new support vector and the new device feature vector. Once the new context vector is generated, the new context vector may be provided to the machine learning engine 106 for predicting troubleshooting solution for the new device model. In one example, the machine learning engine 106 may update the mapping table to include the new context vector and the new device model.

Once the machine learning engine 106 is trained, the system 102 or a similar system may be implemented for interaction with the users of the electronic device 202. In one example, when a user of the electronic device 202 encounters an issue with the electronic device 202, the user may provide a troubleshoot query to the query engine 104. In on example, the troubleshoot query may indicate a device model of the electronic device and an issue with the electronic device. For example, a user of electronic device 202-1 may encounter an issue that the electronic device 202-N, i.e., an imaging device, “cannot print black color objects”. The user may then provide the troubleshoot query indicating the device model of the imaging device 202-1 and the issue faced by the user to the query engine 104.

The query engine 104 may identify a context vector identified corresponding to the device model based on the mapping table and the device model. As previously described, the context vector describes a relationship between the device model, a plurality of troubleshoot support documents, and the functional features of the device model. Table 1 provided below illustrates an example mapping table corresponding to a plurality of device model.

TABLE 1

Electronic
[5.88601595e−01, 4.21660016e−02, −8.55350062e−02]

device 202-1

Electronic
[5.88601595e−01, 4.21660016e−02, −8.55350062e−02]

device 202-2

Electronic
[2.13702135e−02, −2.01120702e−02, 1.63796936e−01]

device 202-N

As illustrated above, a first context vector corresponds to the electronic device 202-1, a second context vector corresponds to the electronic device 202-2, and a third context vector corresponds to the electronic device 202-N. Referring to the troubleshoot query, the query engine 104 may identify the first context vector in order to provide the user with the solution for query “cannot print black color”.

Subsequently, the machine learning engine 106 may determine a troubleshooting solution for the issue based on the issue and the context vector corresponding to the device model. In one example, the machine learning engine 106 may identify a troubleshoot support document describing the troubleshooting solution for the issue from among the plurality of troubleshoot support documents, based on the issue and context vector to provide the troubleshooting solution. In another example, the machine learning engine 106 may analyze a plurality of case logs corresponding to the issue to identify the troubleshooting solution previously provided to a user for the same issue and the similar device model. The machine learning engine 106 may accordingly provide the troubleshooting solution to the user.

FIGS. 3, 4, and 5 illustrate example methods 300, 400, and 500, respectively, for providing troubleshooting solutions. The order in which the methods are described is not intended to be construed as a limitation, and any number of the described method blocks may be combined in any order to implement the methods, or an alternative method. Furthermore, methods 300, 400, and 500 may be implemented by processing resource or computing device(s) through any suitable hardware, non-transitory machine-readable instructions, or combination thereof.

It may also be understood that methods 300, 400, and 500 may be performed by programmed computing devices, such as the electronic devices 202 and the system 102, as depicted in FIGS. 1-2. Furthermore, the methods 300, 400, and 500 may be executed based on instructions stored in a non-transitory computer readable medium, as will be readily understood. The non-transitory computer readable medium may include, for example, digital memories, magnetic storage media, such as one or more magnetic disks and magnetic tapes, hard drives, or optically readable digital data storage media. While the methods 300, 400, and 500 are described below with reference to the electronic device 202 and the system 102 as described above; other suitable systems for the execution of these methods may also be utilized. Additionally, implementation of these methods is not limited to such examples.

FIG. 3 illustrates the method 300 for training the machine learning model, according to an example implementation of the present subject matter. At block 302, a context vector may be generated for each of a plurality of device models based at least on a support vector corresponding to the device model. In one example, the context vector may describe a relationship between the device model, a plurality of troubleshoot support documents, and functional features of the device model. The troubleshoot support documents may provide plurality of options and steps to solve the issue faced by the user of the electronic device. In one example, the troubleshoot support manual may include a plurality of pages having the troubleshoot support documents for various device models. In one example, a troubleshoot support document may refer to a page, one or more pages, logs, records, and parts or sections of a page. For instance, a single page of the troubleshoot support manual may include a troubleshoot support document that may provide troubleshooting steps to overcome an issue. In another example, a troubleshoot support document may span multiple pages of the troubleshoot support manual that may provide troubleshooting steps for an issue. The functional features may represent behavioral objective associated with the particular electronic device. Examples of functional features of an imaging device may include, but are not limited to, scanning, faxing, printing, and copying. Similarly, functional features of a network peripheral may include routing and repeating messages.